BEGIN:VCALENDAR VERSION:2.0 PRODID:-//Memento EPFL// BEGIN:VEVENT SUMMARY:Foundation Models Meet Control: Promises\, Pitfalls\, and Open Que stions DTSTART:20260305T110000 DTEND:20260305T120000 DTSTAMP:20260416T193038Z UID:8f3caf0b7203ae3719a5bd3983f220f81829693bc40cbe703b1006bd CATEGORIES:Conferences - Seminars DESCRIPTION:Prof. Simone Formentin\, Associate Professor of Automatic Cont rol Politecnico di Milano\, Italy\nAbstract: \nFoundation models are rapi dly reshaping engineering and science. But what do they truly imply for co ntrol systems?\nAre they merely powerful black-box approximators\, or do t hey introduce a new computational paradigm for modeling\, estimation\, and controller synthesis?\nThis talk revisits foundation models from a contro l-theoretic perspective. We first clarify how they fundamentally differ fr om classical system identification: rather than learning a single dynamica l model\, foundation models learn conditional distributions over families of admissible solutions.\nInference and design are thus performed through conditioning\, rather than through explicit parameter estimation.\nBuildin g on this viewpoint\, we present two research directions that illustrate t heir potential in model-based and data-driven control\, respectively.\nThe first leverages diffusion models to learn and sample from the distributio n of stabilizing controllers for a given plant.\nThe second explores in-co ntext learning for control and state estimation\, showing how transformer- based architectures can generate filters and feedback laws for previously unseen systems.\nLarge-scale pretraining in simulation enables rapid adapt ation with minimal contextual data\, effectively shifting the sim-to-real challenge from model fidelity to distributional coverage\, that is\, from a model-centric to a data-centric paradigm.\nDespite their promise\, found ation models raise fundamental questions for control\, including stability guarantees\, robustness under distribution shift\, computational burden\, and interpretability.\nWe argue that these limitations should not discour age their adoption in feedback systems\, but instead motivate deeper integ ration: control theory can contribute back to foundation models by providi ng system-theoretic analysis tools\, structural parameterizations\, and pr incipled notions of safety and robustness.\n\nBiography:\nProf. Simone Fo rmentin was born in Legnano\, Italy\, in 1984. He received his B.Sc. and M .Sc. degrees cum laude in Automation and Control Engineering from Polite cnico di Milano\, Italy\, in 2006 and 2008\, respectively. In 2012\, he ob tained his Ph.D. degree cum laude in Information Technology through a jo int program between Politecnico di Milano and Johannes Kepler University o f Linz\, Austria. He subsequently held postdoctoral positions at the Swiss Federal Institute of Technology of Lausanne (EPFL)\, Switzerland\, and at the University of Bergamo\, Italy. Since 2014\, he has been with Politecn ico di Milano\, where he is currently an Associate Professor of Automatic Control. From 2020 to 2025\, he served as Chair of the IEEE Technical Comm ittee on System Identification and Adaptive Control. He is currently an As sociate Editor of Automatica and the European Journal of Control His r esearch interests include data-driven modeling and control\, with applicat ions to automotive systems and quantitative finance.\n  LOCATION:ME C2 405 https://plan.epfl.ch/?room==ME%20C2%20405 STATUS:CONFIRMED END:VEVENT END:VCALENDAR