Pierre-Antoine Lequeu, PhD student
I am on my last year as a PhD student at Sorbonne University, Paris, France under the supervision of Benjamin Piwowarksi at the ISIR laboratory. My PhD topic tackles the Fair and Explainable Automatic Summarization of Political Debates, with a focus on large-scale citizens consultations. My PhD project is part of the Democratic Commons initiative, in parternship with SciencesPo university and Make.org.
I am also a teacher assistant at Sorbonne University for bachelor students and at the MIND master Program.
My main research interests are to understand (1) how Large Language Models work and (2) how they affect society, and in particular democracies. I believe these two to be very much related tasks.
- Concerning the former, my current work focuses on how model use positional information, with a long term objective to design more explicit positional encoding methods (See Give It Space!).
- For the latter, I work on AI-aided deliberation, citizen consultations, and collective decision-making. We designed Corpus Clarification as locally-runnable opinions corpora preprocessing method to facilitate their analysis and released GDN-CC-large, a French corpus of 240k preprocessed opinions for the Grand Débat National Consultation (See GDN-CC). In another work, we show that recommender systems significantly distort the opinion landscape when infering participant’s preferences (Under review, see Collective-Centric evaluation).
I would be more than happy to exchange on my (or your) work and collaborate. You can contact me at [last-name]@isir.upmc.fr on which i should answer fairly quickly.
I am also looking for opportunities starting November 2027 as post-doc or research scientist in AI-enhanced democracies, collective decision making, or related fields.
Otherwise, my life interests are music, drawing, and overall anything art :)
News
✨ 02/09/2026 - We just published a new Preprint: Toward Collective-Centric Evaluation of Preference Inference.
✨ 21/08/2026 - Our paper Give it Space! Explicit Disentangling of Positional and Semantic Representations in Encoders has been accepted at EMNLP 2026! see you there :)
✨ 07/07/2026 - I presented our work on The GDN-CC Dataset at ACL in San Diego, Calfifornia!