Embed Days Colloquium 2025

Keynote Speakers

Thierry Poibeau

Research Director at CNRS, Assisting Director of Lattice Laboratory

We welcome Thierry Poibeau as one of our keynote speakers. As a distinguished Research Director at CNRS and Assisting Director of the Lattice laboratory, his work spans computational linguistics, natural language processing, and artificial intelligence. His expertise in text analysis and language technologies brings valuable insights to the embedding research community.

10am-11am: Do Large Language Models Have Any Relevance for Linguistic Theory?

Mehdi Khamassi

Research Director at CNRS

We are glad to have Mehdi Khamassi join us as a keynote speaker. As a Research Director at CNRS, his work focuses on computational neuroscience and robotics. His research brings together neural networks, reinforcement learning, and cognitive architectures, offering unique perspectives on embedding applications in adaptive behavior and decision-making systems.

13:30-14:30: How can we ensure AI system's alignment with human values ?

Giovanni Petri

Professor in the Network Science Institute at Northeastern University London

Giovanni Petri is joining us as a keynote speaker. His research spans the analysis of neuroimaging data and AI systems with topological techniques, the formalization of cognitive control models with tools of statistical mechanics and network theory, and the study of the predictability of socio-technical systems.

16:15-17:15: Balancing Learning, Generalization, and Processing Efficiency

Practical Information

Preliminary Program

9:00 - 9:30

Breakfast

9:30 - 9:45

Welcome & Opening Remarks

Liubov Tupikina and Charles de Dampierre

9:45 - 10:45

Keynote Talk

Thierry Poibeau

Do Large Language Models Have Any Relevance for Linguistic Theory?

10:45 - 11:00

Coffee Break

11:15 - 12:00

Research Presentations & Lightning Talks

Lorenzo Dall'Amico (online) An embedding-based distance for temporal graphs
Marjolaine Ray An incremental clustering for event detection in press and social media
Jean Barré atent structures of intertextuality in french fiction
Benoit de Courson Measuring semantic coherence in French rap lyrics

12:00 - 13:30

Lunch Break

13:30 - 14:30

Keynote Talk

Mehdi Khamassi

How can we ensure AI system's alignment with human values?

14:30 - 14:45

Coffee Break

14:45 - 15:45

Research Presentations & Lightning Talks

Yuri Gardinazzi (online) Persistent topological features in large language models
Chiara Semenzin Investigating acoustic feature representation in dolphin sounds using Wav2Vec2
Linnea Evanson Decoding the language hierarchy in the neural responses of the child's brain using self-supervised LLM embeddings

15:45 - 16:00

Coffee Break

16:15 - 17:15

Keynote Talk

Giovanni Petri

Balancing Learning, Generalization, and Processing Efficiency

17:15 - 18:30

Closing Discussion & Networking Reception

Location

École Normale Supérieure (ENS)

Salle Dussane

45 rue d'Ulm, 75005 Paris, France

View on Google Maps

Metro: Line 7 (Place Monge), RER B (Luxembourg)

Bus: Lines 21, 27, 89

Online Meeting

For those attending remotely, you can join via Zoom:

Join Zoom Meeting

Meeting ID: 93798697609

Please join 10 minutes before the start of the session.

The Embed-Days will be recorded.

Join the Discussion

Share your work in a format that suits you best. The Embed-Days colloquium welcomes various form of work.

Research Talks (10-25 min)

Present your research on embedding methods, their theoretical foundations, or applications. We welcome both completed work and ongoing research that explores novel approaches, theoretical frameworks, or empirical findings in embedding-based analysis.

Lightning Talks (5-15 min)

A format designed for presenting emerging ideas, methodological innovations, or preliminary results. Particularly suited for work that warrant discussion with the research community.

Discussion Sessions (10-20 min)

Animate a discussion on theoretical challenges, methodological approaches, and recent developments in embedding research. These sessions examine both fundamental questions in embedding theory and their practical implications across different domains of application. You can come with either your own questions or papers to be discussed live.

Four Fields, One Goal

Bringing together diverse perspectives on embeddings

Mathematics

Exploring the theoretical foundations of embeddings, from topology to geometric deep learning

AI Explainability

Understanding and interpreting embedding spaces in artificial intelligence systems

Computational Social Sciences

Applying embedding techniques to understand social phenomena and human behavior

Linguistic and neuroscience

Leveraging embeddings for large-scale data analysis and visualization

About

Research Focus

The Embed-Days colloquium is dedicated to the presentation of research papers on embeddings, focusing on four primary fields: Mathematics, AI Explainability, Computational Social Sciences, and Big Data.

Academic Exchange

The Embed-Days aim to facilitate the exchange of ideas and methodologies in embedding research. While located in Paris, we intend to gather the international embedding research community.

Interdisciplinary Approach

We explore the intersections of embedding research across various domains. This approach allows for a broader understanding of embeddings' potential applications and implications.

The Embed-Days Colloquium

Keynote Speakers

Thierry Poibeau

Mehdi Khamassi

Giovanni Petri

Practical Information

Preliminary Program

Breakfast

Welcome & Opening Remarks

Keynote Talk

Coffee Break

Research Presentations & Lightning Talks

Lunch Break

Keynote Talk

Coffee Break

Research Presentations & Lightning Talks

Coffee Break

Keynote Talk

Closing Discussion & Networking Reception

Location

Online Meeting

Join the Discussion

Research Talks (10-25 min)

Lightning Talks (5-15 min)

Discussion Sessions (10-20 min)

Four Fields, One Goal

Mathematics

AI Explainability

Computational Social Sciences

Linguistic and neuroscience

About

Research Focus

Academic Exchange

Interdisciplinary Approach

Our Partners

PSL University

BunkaLab

CNRS

ENS Ulm

French Ministry of Culture

Nokia Bell Labs

Pr[AI]rie Institute

Institut Jean-Nicod

Our Team

Scientific Committee

Liubov Tupikina

Francesco Vaccarino

Sean Kennedy

Giovanni Petri

Ginestra Bianconi

Organizers

Charles de Dampierre

Valentin Thouzeau

Louis-Marie Lorin

Connect

Stay Updated