8:30 8:40

Opening Remarks

8:40 9:30
Panel

What is a World Model for Robotics?

Tom Silver, Siddharth Srivastava, Sherry Yang

8:40 - 9:00 Panel discussion
9:00 - 9:30 Split panelists and audience into groups to discuss
9:30 10:00
Invited Talk

Evaluating and Improving Physical Agents in a World Model

Sherry Yang · New York University, Google DeepMind

10:00 10:15
Accepted Papers

1st Lightning Talks Session

Watch recording
View detailed schedule
10:00 - 10:02 Causal Neuro-Symbolic World Models for Robotics: A Taxonomy Bridging Interventional Reasoning and Physical Autonomy
10:02 - 10:04 Grounded-Dreamer: Robot Learning from World Model Synthetic Data with Physical Grounding
10:04 - 10:06 DexTouch-WM: Learning Action-Conditioned Tactile World Models from Human Touch for Dexterous Robot Manipulation
10:06 - 10:08 Wild-World: Action-Conditioned RGB-D-Semantic World Modeling for Field Robotics
10:08 - 10:10 Motion as Communication: Plan-Conditioned Diffusion World Models for Negotiation at Unprotected Conflict Points
10:10 - 10:12 Relationally Grounded Latent World Models for Autonomous Driving
10:12 - 10:14 Abstract World Models: Flow Policies as Implicit Action Representations
10:15 11:00
Poster Session Break

1st Poster Session and Coffee Break

10:30 - 11:00 Coffee Break
11:00 11:30
Invited Talk

Building Physical Agents via Structured Representations

Jiajun Wu · Stanford University

Watch recording
11:30 12:00
Invited Talk

Towards Robots that Generate Micro-World Models on Demand

Tom Silver and Yixuan Huang · Princeton University

12:00 12:30
Invited Talk

Learning Relational World Models for Robot Planning

Siddharth Srivastava · Arizona State University

12:30 1:30
Break

Lunch (On Your Own)

1:30 2:00
Invited Talk

Emergence of Action and Object Symbols and Symbolic Rules in Robotics

Emre Ugur · Bogazici University

Watch recording
2:00 2:30
Invited Talk

Towards World Theory Models

Sungjin Ahn · KAIST, New York University

Watch recording
2:30 3:00
Accepted Papers Spotlight

Spotlight Talks

Watch recording
View detailed schedule
2:30 - 2:36 GlanceWAM: Sparse Test-Time Imagination for World-Action Models
2:36 - 2:42 μ0: A Scalable 3D Interaction-Trace World Model
2:42 - 2:48 Bilevel Planning with Learned Symbolic Abstractions from Interaction Data
2:48 - 2:54 CLAP: Cross-Embodiment Video World Models are Zero-Shot Physical Simulators
2:54 - 3:00 Bridging Learned Visual Perception and Symbolic Belief-Space Planning via Probabilistic Grounding
3:00 3:15
Accepted Papers

2nd Lightning Talks Session

Watch recording
View detailed schedule
3:00 - 3:02 iFlax: Neuro-Symbolic Learning for Long-Horizon Task Planning Under Complex Logical Constraints
3:02 - 3:04 STRIPS-WM: Learning Grounded Propositional STRIPS-style World Models from Images
3:04 - 3:06 Option Discovery Using LLM-guided Semantic Hierarchical Reinforcement Learning
3:06 - 3:08 Four Planners, One Building: Shared-Task Benchmarking of Neural and Neurosymbolic World Models, and Three Ways It Misleads
3:08 - 3:10 A Faithful Heading Variable Emerges in Predictive World Models Only Under Partial Observability
3:10 - 3:12 Fires Precisely, Not Early: A Pre-Registered Geodesic Trust Signal for World-Model Rollouts
3:12 - 3:14 Benchmarking and Advancing Physical Reasoning in Large Language Models Through Agents
3:15 4:00
Poster Session Break

2nd Poster Session and Coffee Break

3:30 - 4:00 PM Coffee Break
4:00 4:30
Invited Talk

Planning with Video Based World Models

Yilun Du · Harvard University

Watch recording
4:30 5:20
Panel

Integrating Neural and Symbolic Architectures for Robot World Models

Sungjin Ahn, Yilun Du, Emre Ugur

4:30 - 5:05 Panel discussion
5:05 - 5:20 Audience Q&A
5:20 5:30

Closing Remarks and Best Paper Award Announcement