I am a CNRS research fellow (chargée de recherche) in the MLIA team at ISIR, Sorbonne Université, Paris, since February 2026. My research focuses on natural language processing and machine learning, with three main axes: (1) multimodal models for jointly understanding text, images, and video; (2) multilingual models for better representing under-resourced languages and cultures; and (3) machine reasoning, including mathematical, causal, and visual reasoning.

Before joining CNRS, I was a visiting researcher at UC Berkeley (2025, with Alane Suhr) and a post-doctoral researcher at EPFL in the NLP lab (2023-2025, with Antoine Bosselut), where I worked on multimodal language models and reasoning. Earlier, I was an invited researcher at the Jozef Stefan Institute in Ljubljana (2022), working on multi-task learning and causal discovery, and a postdoc at INRIA Paris in the Almanach team (2021-2022), working on hate speech detection and cross-lingual transfer.

I did my PhD at LISN and Université Paris-Saclay (2018-2021), supervised by Alexandre Allauzen, on computational models of diachronic semantic change. I remain active in the semantic change community, co-organizing the LChange workshop on Computational Approaches to Historical Language Change since 2021.

News:

• 03/2026: Presenting DrivingVQA at EACL 2026 (Rabat)
• 02/2026: Joined ISIR, Sorbonne Université as a CNRS research fellow (chargée de recherche) in the MLIA team.