• Reception to follow in MLH 3. Join us to chat with the speaker and other attendees.

Speaker: Alex Schwing, Ph.D., Associate Professor, University of Illinois at Urbana-Champaign

Abstract
In this talk, I will present recent work focused on looking "behind the seen," i.e., peeking beyond immediately observed data. Specifically, I will cover three domains: holistic environment understanding, forecasting, and optimization. Following a brief primer on generative AI methods and their connections, I will first discuss the benefits of integrating global environment representations with local observations in Large Language Models (LLMs). Second, I will introduce a unified approach to pose and trajectory forecasting that is remarkably simple, easily interfaces with LLMs, and yields compelling results. Finally, I will address optimization by demonstrating how to model dependencies among concurrently unmasked discrete random variables in discrete diffusion models, a crucial step for solving optimization problems using generative AI methods.

Bio
Alex Schwing is an Associate Professor at the University of Illinois at Urbana-Champaign directing a team of talented students on artificial intelligence, generative AI, and computer vision topics. He obtained a PhD in Computer Science from ETH Zurich in 2014 and joined the University of Toronto as a postdoctoral fellow until 2016. His PhD thesis was awarded an ETH medal and his team’s research was awarded an NSF CAREER award. He regularly publishes at top-tier AI conferences, including CVPR and NeurIPS. He served as program co-chair for CVPR 2026. For additional info, please browse to http://alexander-schwing.de.