Laboratoire d'Informatique de Paris Nord

🍃 This week, on 1 April 2025, the LIPN and the LAGA are organising a day of talks on probability and combinatorics, as part of the MathSTIC federation of the Sorbonne-Paris-Nord University.
🤝 The day is co-organised by LAGA's Probability and Statistics team and LIPN's CALIN (Combinatorics, Algorithms and Interactions) team.
❓ For more information on the programme, visit https://lipn.fr/~cb/Seminaires/.
🔗 The presentations will be broadcast on BBB https://bbb.lipn.univ-paris13.fr/b/ban-kpe-96w

🤝 Rafik Aguech is a professor at the University of Monastir. He specializes in probability theory and stochastic processes. His work focuses on random walks, urn models, and reinforced processes, with applications in statistical physics and complex systems.
👨‍🏫 We're thrilled to welcome him as an invited professor at the @LipnLab between March 13th 2025 and April 6th 2025. He will be an invited speaker at the MathStic Days on April 1st.

📃 The paper "PACT: Pruning and Clustering-Based Token Reduction for Faster Visual Language Models", by Mohamed Dhouib (LIX), Davide Buscaldi (@LipnLab), Sonia Vanier (LIX) and Aymen SHABOU (Crédit .Agricole) has been accepted at the CVPR 2025 conference.
❓ PACT is a novel method that reduces the inference time and memory usage of Visual Language Models, contributing to the community's efforts to make multimodal LLMs usage more frugal.

📃 The publication "Linking Industry Sectors and Financial Statements: A Hybrid Approach for Company Classification" by Guy-Stephane Waffo Dzuyo (Loria - Forvis Mazars), Gaël Guibon @LipnLab, Christophe Cerisara (Loria - CNRS) and Luis Belmar Letelier (Forvis Mazars) was presented last week at AAAI conference in Philadelphia.
❓ In this work, the authors tackle company classification from financial statements and show that textual representation through language models yields strong classification performance, which is also explainable compared to numerical ones.

🎉 Davide Buscaldi will defend his Habilitation à diriger des recherches (HdR) on ‘Improving access to Scientific Literature with Knowledge Graphs and Large Language Models’.
📆 The defence will occur on Tuesday 18 March at 10 am, in room B107 at @LipnLab.

LONG LECTURES
- Amir Yehudayoff (University of Copenhagen & Technion-IIT) on Communication complexity
- Heiko Röglin (University of Bonn) on Complexity beyond worst-case analysis
- Antonina Kolokolova (Memorial University of Newfoundland) on Metacomplexity
- Richard Kueng (Johannes Kepler University Linz) on Quantum computation and complexity

SHORT LECTURES

- Tali Kaufman (Bar-Ilan University) on High dimensional expanders
- Akitoshi Kawamura (Kyoto University) on Higher-order complexity

☀️ The "Complexity as a Kaleidoscope" research school will be held at the CIRM from 31 March to 4 April 2025.
🧑‍🤝‍🧑 It is co-organized by T. Seiller (@thomasseiller) and S. Perifel, researcher and Full Professor at the @LipnLab.
🔗 Program and information:
https://conferences.cirm-math.fr/3225.html

❓ A peer effect can be defined as an effect resulting from interactions between individuals in the same group. In this paper, we compare the results obtained by two types of predictive approaches: (1) a linear averaging model and (2) a generalist first-order logic rule mining approach that has been adapted by using specific language biases. The experiments are being carried out on a real dataset from the ANR MEDSPE project, which concerns medical students' choice of a medical speciality when they select their residency position at the end of their 6th year of study. This work is supported by the ANR ERMES.

🎉 Congratulations to Nathalie Pernelle and Céline Rouveirol, who, with other co-authors, have won the best applicative paper award at the ECG 2025 conference !
▶️ The paper, written in french is entitled "Approches prédictives pour l’étude de l’effet de pairs : application à la prédiction de préférence pour une spécialité des étudiants en médecine", and written by
Mustapha Atmani, Nathalie Pernelle, Céline Rouveirol, Noemi Berlin, Magali Dumontet, Mathieu Lambotte, and Carole Treibich.

💔 The Logic and Verification (LoVe) team has reached a critical size and has now separated into two brand-new teams!
Meet:
📍 LoCal (Logique et Calcul), the new Logic team;
https://lipn.univ-paris13.fr/local-logic-and-computation/
📍 SAFER (Safety and Sécurité Analysis via Formal and Efficient veRification), the new Verification team.
https://lipn.univ-paris13.fr/safer-2/

📢 The LIPN is recruiting! Three Assistant Professors and one Full Professor Positions are open at the LIPN.
⏰ Interviews in the spring of 2025, starting date in September 2025.
- One Assistant Professor in Machine Learning
- One Assistant Professor in Combinatorics
- One Assistant Professor in Combinatorial Optimisation
- One Full Professor on any topic represented at LIPN, with a priority for NLP or ML.
🔗 More details: https://lipn.univ-paris13.fr/recrutement/

This week was Science, Culture, and Career Week at the University. To celebrate, a group of undergraduates visited the lab, interacted with researchers, and listened to scientific presentations!