9:00 - 10:00 AM

Registration and breakfast (in person only)

10:00 - 10:15 AM

Welcoming remarks and event kick-off

10:30 - 11:00 AM

Keynote | Nitish Srivastava

Breaking the Data Barrier: Using Simulation to Train Autonomous Driving Robots

Training neural networks for autonomous driving faces a critical data challenge. Most useful learning happens in edge cases, which are infrequent in the real world. When such events do happen, they often provide limited guidance on the correct course of action. Simulation provides a safe space for robots to explore, learn, and refine mental models for driving. However, the gap between simulation and reality has slowed the deployment of these policies in real scenarios. This talk will showcase Vayu's approach to training robot policies in simulation and transferring them effectively to the real world.

11:05 - 11:35 AM

Keynote | Kelsey Allen

Physical reasoning in natural and artificial intelligence

Physical reasoning is ubiquitous. From every-day challenges, like wielding a broom in order to get something from under a couch, to grand challenges in science and engineering, like creating wind turbines or rocket ships, people adaptively reshape their environments to expand what they are capable of. Understanding and transforming the physical world involves a wide range of cognitive capacities including perception, prediction, and planning. This talk will highlight recent work investigating some of these capacities in both humans and machines, and suggest avenues through which we can build more robust and data-efficient machine physical reasoners.

11: 35 AM - 1:00 PM

Lunch (in person only)

1:00 - 1:30 PM

Keynote | Ruslan Salakhutdinov

Multimodal AI Agents 

In recent years, the rise of Large Language Models (LLMs) with advanced general capabilities has accelerated progress toward building language-guided agents capable of performing complex, multi-step tasks, much like human assistants. Developing agents that can perceive, plan, and act autonomously has long been a central goal of artificial intelligence research.

In this talk I will introduce Multimodal AI agents capable of planning, reasoning, and executing actions on the web, that can not only comprehend textual information but also effectively navigate and interact with visual settings. I will present VisualWebArena, a novel framework for evaluating multimodal autonomous language agents, along with an inference-time search algorithm that enables explicit exploration and multi-step planning in interactive web environments. Next, I will demonstrate how an automated data pipeline can facilitate Internet-scale web-agent training by generating web navigation tasks across 150,000 live websites, deploying LLM agents, and assessing their performance. Finally, I will discuss some insights for developing more capable autonomous agents in both digital and physical environments.

1:30 - 1:45 PM

Closing remarks  (virtual audience closes)

1:45 - 2:00 PM

Light refreshments & workshop organization (in person only)

2:00 - 3:00 PM

Breakout Sessions (in person only)

Computer Vision | Leonid Sigal, Evan Shelhamer, April Khademi, Marcus Brubaker, Igor Gilitschenski

This session will spotlight groundbreaking research projects from the Vector research community, offering insights into the latest advancements in computer vision. Furthermore, a panel discussion will explore the future trajectory of the field, including emerging trends, ethical considerations. The conversation will also address overlooked risks in computer vision, raising critical questions about its broader societal impact and potential unintended consequences.

NLP | Frank Rudzicz, Yuntian Deng, Gerald Penn, Hongyang Zhang, Tracy Jenkin, David Duvenaud

 As NLP evolves in 2025, key questions emerge: Have we reached the limits of scaling laws? Do LLMs truly understand language, or are we just fooling ourselves? Are agentic systems and RAG already fading trends? This session will bring together experts to debate these topics and discuss the role of organizations like Vector in shaping the future of AI, as well as overlooked risks in NLP that demand attention.

ML Security & Privacy | Xi He, Reza Samavi, David Lie, David Emerson, Olivia McLaughlin

As machine learning systems become increasingly integrated into critical applications, ensuring their security and privacy is more important than ever. This breakout session will explore key topics such as differential privacy, federated learning, cryptography, and secure and safe machine learning in practice, as well as foster a panel discussion on the challenges and advancements in the field.

AI Safety | David Duvenaud, Nisarg Shah, Silviu Pitis, John Willes, Jenny Bao, David Glukov

ML Theory | Ruth Urner, Sajad Rahmanian Ashkezari, Mohammad Afzali, Nadim Ghaddar

This session will delve into the mathematical foundations of machine learning, exploring theories that advance modern ML practices. Members of the Vector research community will showcase their latest projects, and examine emerging theoretical challenges and their implications for the future of machine learning.

3:00 - 4:30 PM

Poster presentations & networking (in person only)