Workshop Scope

The concept of the "World Model" focuses on how intelligent agents can understand and model the external interactive worlds/environments to improve their decision-making and planning abilities. Over the past two years, the notion of a world model has undergone both conceptual clarification and significant scaling. Moving beyond a loose idea of “models that can imagine the future in the world”, increasingly framed as scalable engines for modeling and simulating dynamic interactions with the world, positioned at the intersection of generative modeling, sequential decision-making, multimodal representation learning, simulation and interaction, causal robustness, and spatial intelligence. In practice, contemporary world models learn composable dynamics, often primarily from large-scale video and related multimodal data, and expose interfaces for prediction, planning, and behavior learning.

The second edition of World Models workshop discussion along a pipeline: (i) understanding and knowledge extraction → (ii) training and evaluation at scale → (iii) cross-modal and control- centric scaling. Systems -level sessions, robotics/Open- World agent case studies, and failure-mode post-mortems will knit these threads together.

• Understanding the World and Extracting Knowledge.

• World Model Training and Evaluation.

• Scaling World Models Predictions Across Language, Vision, and Control.

• World Models in General Domains: Embodied AI, Healthcare, Natural and Social Sciences, and Beyond.

The workshop covers the widest range of World Models topics, including understanding, modelling, as well as scaling with cutting-edge generative AI. We welcome submissions related to the construction, analysis and applications of world models, by using (but not limited to) the following technology: Model-Based Reinforcement Learning, Causality, Sequential Modelling, Simulation of the Environment, Diffusion Models, Video Generation, 3D reconstruction, Spatial Intelligence, Robotics, and Embodied AI etc.

We also encourage submissions from the Natural Sciences (e.g., physics, chemistry, biology) and Social Sciences (e.g., pedagogy, virtual sociology simulation) related to world/environment construction in the science domain to offer attendees a more comprehensive perspective. In summary, topics of interest mainly include, but are not limited to:

• Understanding World Rules: Exploring World Models capture environment dynamics; causality understanding; spatial-temporal modelling; model-based RL; and theoretical foundations for environment simulation and prediction. 

• World model training and evaluation: strengths, limitations, and challenges of current modelling architectures (e.g. Transformers, RNNs, and SSMs), training algorithms (autoregressive training, diffusion modelling, RL, and normalizing flow) and dataset construction. 

• Scaling World Models prediction and generation across language, vision, and control: Investigating how integrating visual, auditory, and textual data improves realism World Models.

• World Models in general domains: Exploring World Models in robotics, AI, healthcare, natural and social sciences, and beyond to improve prediction and decision-making. 

• Benchmark, Dataset, and Demonstration about World Models such as environment simulation.

• ...

Invited Speakers & Panelist

(acceptance order)

Organizing Team

Advisory Board

Support Team