Maxime Darrin

MILA, ILLS, McGill University, Paris-Saclay University

Stablediffused Maxime

I am currently a PhD candidate in information theory and machine learning applied to NLP under the supervision of Pablo Piantanida and Jackie CK Cheung. I am pursuing a joint PhD programm between the ILLS, Paris-Saclay University and MILA, McGill University. I previously completed a Master’s degree in applied mathematics for machine learning from Sorbonne Université and a Master’s degree in fundamental computer science from ENS de Lyon.

I am also interested in the broader impact and deployement of AI models in society. I have got both a master’s degree in philosophy of science from Paris 1 Panthéon-Sorbonne (La Sorbonne) and an engineering degree with a specialization in Public Affairs and innovation from Les Mines Paris. My thesis deals with the notion of robustness of learnt model and its impact on AI regulation.

Research interests

  • Robustness of text generation
  • OOD detection and abstention mechanisms
  • Communication through language models and collaborative reasoning


Jun 10, 2024 When is an Embedding Model More Promising than Another? [abs] [pdf] (Preprint)
Jun 10, 2024 COSMIC: Mutual Information for Task-Agnostic Summarization Evaluation [abs] [pdf] has been accepted at ACL’2024
Jun 10, 2024 GLIMPSE: Pragmatically Informative Multi-Document Summarization for Scholarly Reviews [abs] [pdf] has been accepted at ACL’2024
Feb 10, 2024 Unsupervised Layer-wise Score Aggregation for Textual OOD Detection [abs] [pdf] has been accepted at AAAI’2024.
Oct 10, 2023 My paper RainProof: An Umbrella to Shield Text Generator from Out-Of-Distribution Data [abs] [pdf] has been accepted at EMNLP’2023.

Latest posts

Selected publications