Welcome & intro

• Slava Voloshynovskiy (UniGE)

Large Models in Low Precision: Power, Limits and Scaling Laws (Compression, searching and information-theoretic insights)

• Dan Alistarh ( ISTA, Austria)

Short-term attention enables long term memorization & Stochastic activations (Compression, searching and information-theoretic insights)

• Hervé Jégou (META)

AI through the Information-Theoretic lens (Compression, searching and information-theoretic insights)

• Slava Voloshynovskiy (UniGE)

TwinTURBO: Semi-Supervised Fine-Tuning of Foundation Models via Mutual Information Decompositions (Compression, searching and information-theoretic insights)

• Quetant Guillaume (UNIGE)

Llama4 in 4 bits

• Brian Pulfer (University of Geneva)

Diffusion LM and Compression

• Pavlo Molchanov (NVIDIA, USA)

The Watermark Awakens (Security)

• Pierre Fernandez (Meta, France)

The power of small language models

• Pavlo Molchanov (NVIDIA, USA)

Handcrafted vs. Learned: A Comparative Evaluation of Digital Watermarking Techniques (Security)

• Vitaliy Kinakh

Attacking and Defending Vision Foundation Models (Security)

• Yury Belousov

TBD (Security)

• Brian Pulfer (University of Geneva)

TBD (Security)

• Seyed Mohsen Moosavi-Dezfooli (Apple, CH)

Round table - industrial perspective of model security and privacy

Boat trip to Yvoire https://maps.app.goo.gl/5xWLTmPKiUcsABdm6

Boat trip -Return to Geneva

Generative World Models for Sample-Efficient Reinforcement Learning (GenAI and World Model)

• Eloi Alonso
• Vincent Micheli

Energy-Based Transformers and the Future of Scaling (GenAI and World Model)

• Alexi Gladstone (University of Illinois Urbana-Champaign/META, the USA)

Mini-Bang Detection and Interpretation: Benchmarking Reasoning on Emergent and Open-Ended Dynamics (GenAI and World Model)

• Andrey Ustuzanin (Constructor University, Germany)

Improving mid-resource language performance in open LLMs (GenAI and World Model)

• Hanna Yukhymenko (Hugging Face, CH)

TBD

• Yury Belousov

Binary Diffusion Probabilistic Model

• Vitaliy Kinakh

Discussion & Brainstorm & Free time

DINOv3 (Vision Foundation Models \& Reasoning)

• Federico Baldassarre (Meta, France)

How perception got solved (albeit unsatisfyingly)

• Lucas Beyer
• Xiaohua Zhaị

Reasoning by thinking compositionally (Vision Foundation Models \& Reasoning)

• Emmanuel Abbe (EPFL / Apple, CH)

World Models for Code (Vision Foundation Models \& Reasoning)

• Taco Cohen (Meta, the UK)

The free transformer (Vision Foundation Models \& Reasoning)

• François Fleuret (University of Geneva /Meta, CH/France)

Vision-language models: in-context learning and adaptation for radio-astronomy (Vision Foundation Models \& Reasoning)

• Mariia Drozdova (Universite de Geneve (CH))

Radio Astronomy in the Era of Large Models (Vision Foundation Models \& Reasoning)

• Erica Lastufka (University of Geneva, CH)

Summary/Conclusion/Outlook

• Slava Voloshynovskiy (UniGE)