World Model Planning for Agents

How this thread developed

1. June 2026

   VLMs improve spatial reasoning by actively imagining novel views with a world simulator

   Demonstrates active imagination of novel views using a world simulator to improve spatial reasoning.

2. 2 further papers

   Aug 2026 · University of Washington

   Reusable latent trajectories improve world-model control while cutting planning time

   Demonstrates that reusable latent trajectories improve world-model control, reducing planning time while maintaining performance.

   released code

3. Aug 2026 · University of Kurdistan Hewlêr, Nanyang Technological University

   Physiological grounding makes clinical language-model recommendations safer in controlled tests

   Uses a physiological world model to ground clinical recommendations, improving safety and constraint adherence.

4. 5 further papers

   Sept 2026 · Mohamed bin Zayed University of Artificial Intelligence, University of Science and Technology of China

   Deep research agents struggle with long, multimodal evidence chains

   Requires planning over many intermediate conclusions, highlighting limitations in current agent reasoning.

   released code

5. 3 further papers

   Sept 2026

   Touch-aware world models improve contact-rich dexterous robot manipulation

   Extends world models with touch-awareness for contact-rich manipulation, improving planning accuracy in dexterous robot tasks.

6. 1 further paper

   Sept 2026 · Gaoling School of Artificial Intelligence, Renmin University of China

   Editing agent reasoning history boosts long-horizon task performance

   Proposes AEWM, a world model that edits agent reasoning history to improve long-horizon task performance.

7. Sept 2026 · HKUST(GZ), CUHK

   Visual action rehearsal helps language models control robot manipulation

   Introduces World Action Agent, a visual workspace for rehearsal and correction of robot actions, achieving high success on LIBERO-Pro.

8. Sept 2026 · Fondazione Bruno Kessler, Carnegie Mellon University

   Coding agents synthesize robot planners that generalize to unseen instances

   Demonstrates that coding agents can generate reusable robot planners that generalize across many unseen task environments.

9. 1 further paper

   Sept 2026 · Virginia Tech, University of Wisconsin Madison

   Structural guidance improves language models’ long-horizon reasoning under sparse rewards

   Directly addresses long-horizon planning biases in LLMs, proposing topological guidance as a method to improve world-model-based reasoning.

   released code

10. Sept 2026 · University of Chinese Academy of Sciences, Institute of Automation, Chinese Academy of Sciences

    GUI world models fail when hidden state is omitted from predictions

    StateAliasBench isolates a systematic failure of GUI world models that omit hidden state from transition-relevant predictions.

11. 1 further paper

    Oct 2026 · Tsinghua University, Manifold AI

    Aiming at observed intermediate targets beats final-goal scoring in frozen world model planning

    Introduces intermediate target scoring to improve long-horizon planning in frozen visual world models.

12. Oct 2026 · New York University, AMI Labs

    Compressed world models make zero-shot driving plans far faster

    Evaluates compressed latent JEPA as a driving world model for fast zero-shot planning without training a policy.

    released code

12 of 25 papers shown