People

PhD students

Lisa Bouger — PhD (2026–2029), CIFRE Thales. LLM safety; interpretable guardrails. Co-supervisor (50% with Y. Teglia, Thales).
Maxence Lasbordes — Predoc (2025–2026), PhD (2026–2029), Scribe / CIFRE LightOn. Post-training, alignment, synthetic data. Co-supervisor (50%, with I. Poli, LightOn).
Théo Lasnier — Predoc (2025), PhD (2026–2029), Code Common / Scribe / Prairie-PSAI. LLM-generated code evaluation, interpretability, sparse autoencoders. Supervisor (100%).
Romain Froger — PhD (2025–2028), CIFRE Meta. LLM agents: reasoning, planning, tool use. Co-supervisor (50%, with T. Scaliom, Meta).
Sofia de Témiolles — PhD (2025–2028), AeX SaLM. Bias in NLP evaluation; fairness, context-sensitive models. Co-supervisor (50%, with J.-P. Cointet, Sciences Po).
Wissam Antoun — Predoc (2022–2023), PhD (2023–2026), Counter / Ministry of Interior. Detection of LLM-generated content; backdoored LLMs. Co-supervisor (80%, with B. Sagot).
Arij Riabi — Predoc (2020–2021), PhD (2021–2024), Prairie / Counter. Multilingual radical content detection; annotator bias; privacy-preserving datasets. Supervisor (90%, with L. Romary).

Anna Mosolova — Postdoc (2025–2026), BPI Scribe. Post-training and evaluation of domain-specialized LLMs and agents. Supervisor (100%).
Yannis Karmim — Postdoc (2025–2027), DRI Funding (Inria Chile). Structured data inclusion in LLMs; cultural bias evaluation and mitigation. Co-supervisor (50%, with L. Marti, Inria Chile).
Virginie Mouilleron — Research Engineer (2022–2027), Counter / Scribe. Corpus annotation; bias evaluation; adversarial datasets. Supervisor (100%).

Arij Riabi — Predoc (2020–2021), PhD (2021–2024), Prairie / Counter. Multilingual radical content detection; annotator bias; privacy-preserving datasets. Supervisor (90%, with L. Romary). pdf
Benjamin Muller — PhD (2018–2022), Parsiti / SoSweet / Counter. Interpretability; cross-lingual transfer in multilingual LLMs. Co-supervisor (80%, with B. Sagot). pdf
José Carlos Rosales Núñez — PhD (2017–2023), Postdoc (2024), Parsiti. Robust machine translation of noisy user-generated content; multitask classification. Co-supervisor (40% PhD, with G. Wisniewski; 100% Postdoc). pdf
Ghazi Fheli — PhD (2019–2023), Parsiti. Interpretability of neural models; disentangling syntax and semantics. Co-supervisor (40%, with J. Le Roux). pdf
Corentin Ribeyre — PhD (2012–2016), MENRT. Data-driven deep-syntax parsing; graph-based representations. Co-supervisor (45%, with E. de la Clergerie, L. Danlos). pdf