mlsanityData to 14 Oct 2026

GeoWorld: Geometric World Models

arXiv 2602.23058 ↗First posted 26 Feb 2026cs.CV cs.ROcited by 6

On the map: World models → Latent predictive world models

Abstract (arXiv)

Energy-based predictive world models provide a powerful approach for multi-step visual planning by reasoning over latent energy landscapes rather than generating pixels. However, existing approaches face two major challenges: (i) their latent representations are typically learned in Euclidean space, neglecting the underlying geometric and hierarchical structure among states, and (ii) they struggle with long-horizon prediction, which leads to rapid degradation across extended rollouts. To address these challenges, we introduce GeoWorld, a geometric world model that preserves geometric structure and hierarchical relations through a Hyperbolic JEPA, which maps latent representations from Euclidean space onto hyperbolic manifolds. We further introduce Geometric Reinforcement Learning for energy-based optimization, enabling stable multi-step planning in hyperbolic latent space. Extensive experiments on CrossTask and COIN demonstrate around 3% SR improvement in 3-step planning and 2% SR improvement in 4-step planning compared to the state-of-the-art V-JEPA 2. Project website: https://steve-zeyu-zhang.github.io/GeoWorld.

Key facts. Cited by 6 papers in the map (data through 14 Oct 2026). Placed in Latent predictive world models, one of 11 directions in World models. Builds on 64 of its 89 references that are in the map. First posted on arXiv 26 Feb 2026.

Builds on

64 of 89 references.

Closest ancestors · 3

Same direction, newest first

  1. 2024Learning and Leveraging World Models in Visual Representation Learning1 Mar
  2. 2024Revisiting Feature Prediction for Learning Visual Representations from Video15 Feb
  3. 2023Self-Supervised Learning from Images with a Joint-Embedding Predictive Architecture19 Jan

Other references · 61

  1. 2022Training language models to follow instructions with human feedbackcited by 15,971
  2. 2017Proximal Policy Optimization Algorithmscited by 15,215
  3. 2022Chain-of-Thought Prompting Elicits Reasoning in Large Language Modelscited by 12,459
  4. 2025Qwen3 Technical Reportcited by 8,820
  5. 2025Gemini 2.5: Pushing the Frontier with Advanced Reasoning, Multimodality, Long Context, and Next Generation Agentic Capabilitiescited by 3,796
  6. 2024π0: A Vision-Language-Action Flow Model for General Robot Controlcited by 2,407
  7. 2017The Kinetics Human Action Video Datasetcited by 1,909
  8. 2025π0.5: a Vision-Language-Action Model with Open-World Generalizationcited by 1,761
  9. 2025InternVL3.5: Advancing Open-Source Multimodal Models in Versatility, Reasoning, and Efficiencycited by 1,510
  10. 2017The "something something" video database for learning and evaluating visual common sensecited by 1,103
  11. 2019HowTo100M: Learning a Text-Video Embedding by Watching Hundred Million Narrated Video Clipscited by 950
  12. 2024DROID: A Large-Scale In-The-Wild Robot Manipulation Datasetcited by 852
38 more
  1. 2025V-JEPA 2: Self-Supervised Video Models Enable Understanding, Prediction and Planningcited by 809
  2. 2017Poincar'e Embeddings for Learning Hierarchical Representationscited by 684
  3. 2025Self Forcing: Bridging the Train-Test Gap in Autoregressive Video Diffusioncited by 625
  4. 2018Hyperbolic Neural Networkscited by 402
  5. 2019Hyperbolic Graph Convolutional Neural Networkscited by 402
  6. 2019COIN: A Large-scale Dataset for Comprehensive Instructional Video Analysiscited by 318
  7. 2025MAGI-1: Autoregressive Video Generation at Scalecited by 276
  8. 2025LongLive: Real-time Interactive Long Video Generationcited by 265
  9. 2025SkyReels-V2: Infinite-length Film Generative Modelcited by 229
  10. 2019Cross-task weakly supervised learning from instructional videoscited by 199
  11. 2025Self-Forcing++: Towards Minute-Scale High-Quality Video Generationcited by 192
  12. 2024Autoregressive Video Generation without Vector Quantizationcited by 186
  13. 2024Predictive Inverse Dynamics Models are Scalable Learners for Robotic Manipulationcited by 161
  14. 2023Structured World Models from Human Videoscited by 147
  15. 2022The Debate Over Understanding in AI's Large Language Modelscited by 142
  16. 2023Hyperbolic Image-Text Representationscited by 136
  17. 2024Loong: Generating Minute-level Long Videos with Autoregressive Language Modelscited by 99
  18. 2019Procedure Planning in Instructional Videoscited by 92
  19. 2024Compositional Entailment Learning for Hyperbolic Vision-Language Modelscited by 74
  20. 2022Hyperbolic Contrastive Learning for Visual Representations beyond Objectscited by 61
  21. 2021Procedure Planning in Instructional Videos via Contextual Modeling and Model-based Policy Learningcited by 53
  22. 2022P3IV: Probabilistic Procedure Planning from Instructional Videos with Weak Supervisioncited by 52
  23. 20253D-R1: Enhancing Reasoning in 3D VLMs for Unified Scene Understandingcited by 47
  24. 2019Imitation Learning from Observations by Minimizing Inverse Dynamics Disagreementcited by 46
  25. 2021PlaTe: Visually-Grounded Planning with Transformers in Procedural Taskscited by 45
  26. 2025InfinityStar: Unified Spacetime AutoRegressive Modeling for Visual Generationcited by 45
  27. 2025VideoWorld: Exploring Knowledge Learning from Unlabeled Videoscited by 43
  28. 2024Hand-Object Interaction Pretraining from Videoscited by 39
  29. 2025Nav-R1: Reasoning and Navigation in Embodied Scenescited by 38
  30. 2025VLA-R1: Enhancing Reasoning in Vision-Language-Action Modelscited by 38
  31. 2023PDPP: Projected Diffusion for Procedure Planning in Instructional Videoscited by 36
  32. 2025Planning with Reasoning using Vision Language World Modelcited by 36
  33. 2022Learning Iterative Reasoning through Energy Minimizationcited by 32
  34. 2023Inverse Dynamics Pretraining Learns Good Representations for Multitask Imitationcited by 31
  35. 2023Pretrained Language Models as Visual Planners for Human Assistancecited by 28
  36. 2022Hyperbolic Deep Reinforcement Learningcited by 25
  37. 2020An Inductive Bias for Distances: Neural Nets that Respect the Triangle Inequalitycited by 24
  38. 2024SCHEMA: State CHangEs MAtter for Procedure Planning in Instructional Videoscited by 24

50 most cited of 61 All 61 (JSON)

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2026: 6.

6 papers.

  1. 2026LiveWorld: Simulating Out-of-Sight Dynamics in Generative Video World Modelscited by 28
  2. 2026MWM: Mobile World Models for Action-Conditioned Consistent Predictioncited by 7
  3. 2026Orca: The World is in Your Mindcited by 6
  4. 2026Recovering Physical Dynamics from Discrete Observations via Intrinsic Differential Consistencycited by 1
  5. 2026How You Move Tells What You'll Do: Trajectory-Conditioned Egocentric Predictioncited by 1
  6. 2026HSG: Hyperbolic Scene Graphcited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2026ProWorld: Progress-Aware Hyperbolic World Models for Long-Horizon Visual Goal Reaching
  2. 2026EpicWorldModel: Exploration-driven Planning with Latent World Models
  3. 2020World Model as a Graph: Learning Latent Landmarks for Planning
  4. 2026SCALE: State-Calibrated Latent Embeddings for JEPA Planning in the Right Geometry
  5. 2026The Global Neural World Model: Spatially Grounded Discrete Topologies for Action-Conditioned Planning
  6. 2026Sampling-based Kinodynamic Planning with World Models
  7. 2026Beyond the Next Step: Variable-Length Latent World Models for Long-Horizon Planning
  8. 2026FF-JEPA: Long-Horizon Planning in World Models with Latent Planners