AI Seminar Cycle

A monthly seminar series connecting students, PhD candidates and researchers with leading voices in artificial intelligence.

The AI Seminar Cycle is organized by Hi! PARIS, in collaboration with the ELLIS program on Theory, Algorithms, and Computations of Modern Learning Systems and the Paris ELLIS Unit. 
This partnership aims to broaden the visibility of the seminar among the European AI research community.

More about ELLIS

The series offers the Hi! PARIS and broader AI research community the opportunity to discover emerging research topics, engage directly with leading researchers and explore new perspectives across the field.

Most sessions are held online, making the seminar accessible to a broad academic community in France and across Europe.

2026–2027 Seminar Program

Lorenzo Baraldi

Talk title: Inside Multimodal Foundation Models: Representations as a Lever for Control, Safety and Grounding

Wednesday, October 7 (11AM – 12PM) – Online (Register)

Personal Webpage
Jes Frellsen

Talk title and abstract coming soon.

Research areas include probabilistic machine learning, generative AI and uncertainty quantification.

Wednesday, November 4 (11AM – 12PM) – Online (Register)

Personal Webpage
Angela Yao (University of Singapore)

Talk title: When Generators Condition on Themselves – Exposure Bias in Visual Autoregressive Models

Autoregressive models are a strong alternative to diffusion for image and video generation, but generating step-by-step makes them vulnerable to exposure bias, where early errors shape subsequent outputs. This talk presents strategies to reduce exposure bias across the pipeline, including tokenizer and generator regularization, post-training pixel-level alignment, training-free guidance on generated history, and reinforcement learning for long-horizon consistency in video world models.

Monday, November 30 (11AM – 12PM) – Online (Register)

Personal Webpage

Talk title and abstract coming soon.

Research areas include human-centered AI, trustworthy AI and explainable AI.

Wednesday, April 14 (11AM – 12PM) – Online (Register)

Personal Webpage

More sessions will be announced throughout the academic year.

About the Seminar Cycle

The AI Seminar Cycle is designed primarily for students, PhD candidates and researchers, while remaining open to the broader AI research community.

Each session offers an opportunity to:

  • discover research topics beyond one’s immediate field;

  • hear directly from researchers working on current and emerging AI challenges;

  • exchange ideas through questions and discussion;

  • strengthen connections across the European AI research community.

Sessions generally last one hour, including the presentation and discussion with the audience.

Previous Editions

2025-2026

Discover the speakers, topics and sessions from the 2025–2026 AI Seminar Cycle.

→ The 2025–2026 program and past sessions:

  • Xiao-Li Meng, Harvard University | No Free Lunch: “From a Simultaneous (Machine) Learning Impossibility to Heisenberg Uncertainty Principle” | October 2, 2025

  • Elisa Ricci, University of Trento, Hi! PARIS SAB Member | No labels, no training: Leveraging Language for Detecting Anomalous Events in Videos | November 5, 2025

  • Spyros Gidaris, Valeo AI | Latent Representations for Better Generative Image Modeling | December 10, 2025

  • Oriane Siméoni, Meta | Research Session on Vision | February 4, 2026

  • Patrick Perez, Kyutai | Small talk is harder than it sounds | March 4, 2026

  • Francis Bach, Inria | A spectral framework for closed-form relative density estimation | June 3, 2026

Previous editions of the seminar will continue to be archived here as the series develops.

Scientific Committee