Events

CILVR Seminar: Ensuring Reliability and Understanding Theoretical Limits of Foundation Models

Speaker: Karen Ullrich

Location: 60 Fifth Avenue, Room 204

Date: Wednesday, September 17, 2025

Why do even sophisticated LLMs struggle with simple code completions? Why do we find new jailbreaking attacks every day? And what do these failures tell us about how to build agentic systems?

Bio: Karen Ullrich is a research scientist (s/h) at FAIR NYC and actively collaborates with researchers from the Vector Institute and the University of Amsterdam. Her research focus lies in the intersection of information theory and probabilistic machine learning / deep learning. She previously completed a PhD under the supervision of Professor Max Welling. Prior to that, she worked at the Austrian Research Institute for AI in the Intelligent Music Processing and Machine Learning Group led by Prof. Gerhard Widmer. She studied Physics and Numerical Simulations in Leipzig and Amsterdam.