Scholar
Virginie Do
Google Scholar ID: u3QYKhkAAAAJ
Meta
Citations & Impact
Citations
11,282
H-index
14
i10-index
15
Publications
20
Co-authors
0
Publications
Resume (English only)
Academic Achievements
- Published multiple papers on fairness and recommender systems, including 'Robust LLM Safeguarding via Refusal Feature Adversarial Training' (ICLR 2025), 'The Llama 3 Herd of Models' (arXiv 2024), 'Group fairness without demographics using social networks' (ACM FAccT 2023), etc. Received several awards, such as the 2024 AFIA Best Dissertation Award, 2023 French Young Talent Award - L’Oréal-UNESCO For Women in Science, 2022 AAAI Outstanding Paper Award, etc.
Research Experience
- Currently an AI Researcher at Meta, core contributor to ARE project and the Gaia2 benchmark. Published papers in several international conferences such as ICLR, NeurIPS, etc.
Education
- Completed her PhD at PSL University (Université Paris Dauphine) and Meta (FAIR) in 2023, advised by Nicolas Usunier (FAIR), Jamal Atif, and Jérôme Lang (PSL University, CNRS). Her PhD focused on fairness in recommender systems, blending economic concepts with convex optimization and bandit techniques. Prior to her PhD, she received an MSc in Social Data Science from the University of Oxford, an MSc/BSc in Applied Mathematics from Ecole Polytechnique (Diplôme d'Ingénieur), and a BA in Philosophy from Université Paris Nanterre.
Background
- AI Research Scientist, focusing on LLM agents and alignment, with an interest in dynamic environments for evaluating and training LLM agents with RL.
Miscellany
- Co-chaired the workshop 'Responsible Decision-Making in Dynamic Environments' at ICML 2022 in Baltimore, and served as a reviewer for several international conferences.
Co-authors