Key Dates
The planned dates are as follows (all times are UTC-12h, aka “Anywhere on Earth (AOE)”):
We offer a two-stage submission process, with Stage 2 following Stage 1, Submission Site: OpenReview.
Stage 1 Submission and Review Date:
Stage 1 Submission Deadline: 11:59pm, February 10th, 2025 AOE 0:00am, February 14th, 2025 AOE
Stage 1 Review Start: February 14th, 2025
Stage 1 Review End: February 28th, 2025
Stage 2 Submission and Review Date:
Stage 2 Submission Open: February 14th, 2025
Stage 2 Submission Deadline: 11:59pm, February 26th, 2025 AOE
Stage 2 Review Start: February 27th, 2025
Stage 2 Review End: March 5th, 2025
Meta Review: March 3rd-5th, 2025
Notification: March 5th, 2025 AOE
Camera Ready: April 10th AOE
Venue Start Date: April 28th, 2025, Singapore EXPO Peridot 201 & 206
Subject Areas
We welcome submissions related to the construction, analysis and applications of world models, such as Model-Based Reinforcement Learning, Causality, Sequential Modelling, Simulation of the Environment, Diffusion Models, Video Generation, Foundation World Models, 2D to 3D, 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 how World Models capture environment dynamics; World Models by causal understanding, spatial-temporal modelling, model-based RL; and theoretical foundations for 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 embodied AI, robotics, automatic driving, AI, healthcare, natural and social sciences, and beyond to improve prediction and decision-making.
Benchmark, Dataset, and Demonstration about World Models.
...
For a detailed summary of our main focus areas, please visit the front page. Contributions may vary in length from 2 to 9 main pages (Tiny Paper: 2-5 main pages; Full Paper: at most 9 main pages) and unlimited additional pages of reference and appendix. formatted in accordance with ICLR 2025 paper format. We also welcome shorter submissions, particularly those showcasing proof-of-concept work, including live demonstrations, code repositories, or blog posts.
Submission Instruction
The paper submission deadlines are:
Stage 1: 11:59pm, February 10th, 2025 AOE.
Stage 2: 11:59pm, February 26th, 2025 AOE.
Papers must be submitted using the workshop submission system at: OpenReview.
Double-Blind Review Policy:
The review process will be conducted under a double-blind protocol. Authors are responsible for ensuring that their submissions are fully anonymized. This includes omitting author names, affiliations, and acknowledgements from the submission. Authors should avoid disclosing any identifying information, including in supplementary materials.
Dual-Submission Policy:
Submissions to this workshop are non-archival, allowing for the inclusion of ongoing or unpublished work. We also accept submissions of papers currently under review at other venues. However, work that has already been published or accepted for publication elsewhere will not be considered (including ICLR 2025). This policy on dual submissions remains in effect throughout the entire review process.
Transparency:
We will only publicly share the list of accepted papers and their corresponding articles. The review process and review content will not be made public.
Contact:
Please email worldmodels.workshop.25@gmail.com if you have questions.
DeepReviewer Collaboration:
We are pleased to announce that, as part of an experimental initiative, we are collaborating with the DeepReviewer team in the review process. Through a locally deployed DeepReviewer-14B model, we will provide constructive feedback on official reviews while ensuring complete data privacy. DeepReviewer's role is limited to gathering evidence from papers to support or challenge reviewer opinions and offering actionable feedback to reviewers. It is important to note that DeepReviewer-14B model will not directly participate in the review process or provide opinions to authors. This collaboration aims to enhance the quality and consistency of the review process while maintaining the integrity of human judgment.