Nazanin Shafiabadi

Nazanin Shafiabadi

PhD Researcher
ISIR, Sorbonne University

About Me

I am a PhD Researcher in the MLIA team at the ISIR lab (Sorbonne University), working under the supervision of François Yvon. My work bridges Artificial Intelligence and the social sciences. My primary research investigates the expression of bias in multilingual language models—specifically, how these biases vary across languages and how machine translation systems can introduce or amplify cultural distortions in subjective content, such as stance or opinion.

I am currently funded by Bpifrance as a member of the "AI For Democracy Democratic Commons" global research program. In this context, I study the application and impact of AI systems within democratic processes. Ultimately, my goal is to develop language technologies that preserve semantic and pragmatic invariance across languages, ensuring that AI is equitable, transparent, and culturally aware.

Research Interests

Multilingual NLP Evaluating and achieving semantic and pragmatic invariance across languages and translation systems.
AI Fairness & Bias Mitigation Identifying how translation systems introduce or amplify social, cultural, and subjective biases.
AI for Democracy Exploring the safe, equitable usage of AI systems and LLMs within democratic processes.
LM Interpretability Understanding lexical acquisition and the emergence of language abilities in neural LMs during pre-training.

Publications

UNDER REVIEW

Toward Collective-Centric Evaluation of Preference Inference for Participatory Democracy

Pierre-Antoine Lequeu, Salim Hafid, Paul Lerner, Nazanin Shafiabadi, Laurène Cave, David Mas, Jean-Philippe Cointet, Benjamin Piwowarski, François Yvon

AAAI 2027 (Artificial Intelligence for Social Impact Track)

Paper
ACCEPTED

Data Shapes What, Scale Shapes How: Lexical Learning Dynamics in LLMs

Nazanin Shafiabadi, Guillaume Wisniewski

EMNLP 2026

Paper (Coming Soon)
PUBLISHED

Assessing the import of morphology in gender assignment: A computational study of French

Nazanin Shafiabadi, Nina Nusbaumer, Olivier Bonami

Lexique Journal 2026

Paper
PUBLISHED

Biases in Translation: Assessing Opinion Distortion in Machine Translated Texts

Nazanin Shafiabadi, François Yvon

LREC 2026

Paper Code
PUBLISHED

Beyond Surprisal: A Dual Metric Framework for Lexical Skill Acquisition in LLMs

Nazanin Shafiabadi, Guillaume Wisniewski

COLING 2025 · Winner of the Best Short Paper Award

Paper Code Award

Presentations & Posters

POSTER

Biases in Translation: Assessing Opinion Distortion in Machine Translated Texts

Nazanin Shafiabadi, François Yvon

LREC 2026

View Poster
POSTER

Anti-surprise : Une métrique complémentaire pour évaluer l’apprentissage lexical des (grands) modèles de langue

Nazanin Shafiabadi, Guillaume Wisniewski

TALN 2025

View Poster
ORAL

Beyond Surprisal: A Dual Metric Framework for Lexical Skill Acquisition in LLMs

Nazanin Shafiabadi, Guillaume Wisniewski

COLING 2025

View Slides
TALK

Assessing the import of morphology in gender assignment: A computational study of French

Nazanin Shafiabadi, Nina Nusbaumer, Olivier Bonami

LLcD 2024

View Slides

Experience

Research Intern

Apr – Aug 2024

Laboratoire de Linguistique Formelle (LLF) · Paris, France

Supervision: Guillaume Wisniewski

Investigated the mechanisms by which neural language models (particularly BERT) acquire language abilities during pre-training. Extracted and analyzed learning curves from intermediate checkpoints to examine the progression of lexical and syntactic knowledge acquisition, revealing how token frequency and syntactic complexity influence learning dynamics.

Research Intern

Jun – Jul 2023

Laboratoire de Linguistique Formelle (LLF) · Paris, France

Supervision: Olivier Bonami

Evaluated how major LLMs handle neology. Analyzed whether ChatGPT-3.5 could provide accurate definitions for neologisms coined after its training date, and conversely, whether it could coin accurate terms given the definition of these neologisms.

Education

PhD in Natural Language Processing

ISIR Lab, Sorbonne Université (France) · 2024 – Present

MSc in Computational Linguistics

Université Paris Cité (France) · 2022 – 2024

BA in English Language and Literature

Ferdowsi University of Mashhad (Iran) · 2015 – 2019

Training

ALPS Spring School

Mar – Apr 2026 · Aussois, France

International event dedicated to NLP, Speech Processing, and AI featuring collaborative research discussions with leading experts.

LxMLS Summer School

Jul 2025 · Lisbon, Portugal

Deepening of machine learning, deep learning, and NLP processes with hands-on coding labs at Instituto Superior Técnico.