mlsanityData to 7 Oct 2026

VLA-JEPA: Enhancing Vision-Language-Action Model with Latent World Model

arXiv 2602.10098 ↗First posted 10 Feb 2026cs.RO cs.CVcited by 104

On the map: World models → Embodied and robot world models

also in 1 more
Abstract (arXiv)

Pretraining Vision-Language-Action (VLA) policies on internet-scale video is appealing, yet current latent-action objectives often learn the wrong thing: they remain anchored to pixel variation rather than action-relevant state transitions, making them vulnerable to appearance bias, nuisance motion, and information leakage. We introduce VLA-JEPA, a JEPA-style pretraining framework that sidesteps these pitfalls by design. The key idea is leakage-free state prediction: a target encoder produces latent representations from future frames, while the student pathway sees only the current observation -- future information is used solely as supervision targets, never as input. By predicting in latent space rather than pixel space, VLA-JEPA learns dynamics abstractions that are robust to camera motion and irrelevant background changes. This yields a simple two-stage recipe -- JEPA pretraining followed by action-head fine-tuning -- without the multi-stage complexity of prior latent-action pipelines. Experiments on LIBERO, LIBERO-Plus, SimplerEnv and real-world manipulation tasks show that VLA-JEPA achieves consistent gains in generalization and robustness over existing methods.

Key facts. Cited by 104 papers in the map (data through 7 Oct 2026). Placed in Embodied and robot world models, one of 11 directions in World models. Builds on 83 of its 90 references that are in the map. First posted on arXiv 10 Feb 2026.

Builds on

83 of 90 references.

Closest ancestors · 12

Same direction, newest first

  1. 2026Learning Latent Action World Models In The Wild8 Jan
  2. 2025Motus: A Unified Latent Action World Model15 Dec
  3. 2025StaMo: Unsupervised Learning of Generalizable Robot Motion from Compact State Representation6 Oct
  4. 2025FlowVLA: Visual Chain of Thought-based Motion Reasoning for Vision-Language-Action Models25 Aug
  5. 2025WorldVLA: Towards Autoregressive Action World Model26 Jun
  6. 2025V-JEPA 2: Self-Supervised Video Models Enable Understanding, Prediction and Planning11 Jun
  7. 2025AdaWorld: Learning Adaptable World Models with Latent Actions24 Mar
  8. 2025DyWA: Dynamics-adaptive World Action Model for Generalizable Non-prehensile Manipulation21 Mar
  9. 2024Predictive Inverse Dynamics Models are Scalable Learners for Robotic Manipulation19 Dec
  10. 2024IGOR: Image-GOal Representations are the Atomic Control Units for Foundation Models in Embodied AI17 Oct
  11. 20243D-VLA: A 3D Vision-Language-Action Generative World Model14 Mar
  12. 2023Learning to Act without Actions17 Dec

Other references · 71

  1. 2025Qwen3 Technical Reportcited by 8,195
  2. 2023Visual Instruction Tuningcited by 7,792
  3. 2022Scalable Diffusion Models with Transformerscited by 5,931
  4. 2017Neural Discrete Representation Learningcited by 4,650
  5. 2023RT-2: Vision-Language-Action Models Transfer Web Knowledge to Robotic Controlcited by 2,993
  6. 2024OpenVLA: An Open-Source Vision-Language-Action Modelcited by 2,777
  7. 2024π0: A Vision-Language-Action Flow Model for General Robot Controlcited by 2,297
  8. 2022RT-1: Robotics Transformer for Real-World Control at Scalecited by 2,007
  9. 2025π0.5: a Vision-Language-Action Model with Open-World Generalizationcited by 1,653
  10. 2021Ego4D: Around the World in 3,000 Hours of Egocentric Videocited by 1,530
  11. 2023LIBERO: Benchmarking Knowledge Transfer for Lifelong Robot Learningcited by 1,338
  12. 2025GR00T N1: An Open Foundation Model for Generalist Humanoid Robotscited by 1,167
38 more
  1. 2025SigLIP 2: Multilingual Vision-Language Encoders with Improved Semantic Understanding, Localization, and Dense Featurescited by 1,127
  2. 2017The "something something" video database for learning and evaluating visual common sensecited by 1,092
  3. 2025Fine-Tuning Vision-Language-Action Models: Optimizing Speed and Successcited by 844
  4. 2024DROID: A Large-Scale In-The-Wild Robot Manipulation Datasetcited by 801
  5. 2024Genie: Generative Interactive Environmentscited by 706
  6. 2024PaliGemma: A versatile 3B VLM for transfercited by 679
  7. 2025FAST: Efficient Action Tokenization for Vision-Language-Action Modelscited by 643
  8. 2025SpatialVLA: Exploring Spatial Representations for Visual-Language-Action Modelcited by 502
  9. 2024Evaluating Real-World Robot Manipulation Policies in Simulationcited by 484
  10. 2025CoT-VLA: Visual Chain-of-Thought Reasoning for Vision-Language-Action Modelscited by 440
  11. 2025AgiBot World Colosseo: A Large-scale Manipulation Platform for Scalable and Intelligent Embodied Systemscited by 390
  12. 2025UniVLA: Learning to Act Anywhere with Task-centric Latent Actionscited by 386
  13. 2024Universal Manipulation Interface: In-The-Wild Robot Teaching Without In-The-Wild Robotscited by 373
  14. 2024Prismatic VLMs: Investigating the Design Space of Visually-Conditioned Language Modelscited by 336
  15. 2024Latent Action Pretraining from Videoscited by 302
  16. 2023DreamLLM: Synergistic Multimodal Comprehension and Creationcited by 296
  17. 2024Video Prediction Policy: A Generalist Robot Policy with Predictive Visual Representationscited by 279
  18. 2025LIBERO-Plus: In-depth Robustness Analysis of Vision-Language-Action Modelscited by 278
  19. 2021Bridge Data: Boosting Generalization of Robotic Skills with Cross-Domain Datasetscited by 248
  20. 2024NaVid: Video-based VLM Plans the Next Step for Vision-and-Language Navigationcited by 222
  21. 2024RT-H: Action Hierarchies Using Languagecited by 204
  22. 2025DreamVLA: A Vision-Language-Action Model Dreamed with Comprehensive World Knowledgecited by 203
  23. 2024Uni-NaVid: A Video-based Vision-Language-Action Model for Unifying Embodied Navigation Taskscited by 146
  24. 2024Data Scaling Laws in Imitation Learning for Robotic Manipulationcited by 144
  25. 2025SimpleVLA-RL: Scaling VLA Training via Reinforcement Learningcited by 141
  26. 2025VLA-Adapter: An Effective Paradigm for Tiny-Scale Vision-Language-Action Modelcited by 140
  27. 2025RoboBrain: A Unified Brain Model for Robotic Manipulation from Abstract to Concretecited by 134
  28. 2024ShapeLLM: Universal 3D Object Understanding for Embodied Interactioncited by 133
  29. 2025Unified Vision-Language-Action Modelcited by 128
  30. 2025GraspVLA: a Grasping Foundation Model Pre-trained on Billion-scale Synthetic Action Datacited by 120
  31. 2025ChatVLA: Unified Multimodal Understanding and Robot Control with Vision-Language-Action Modelcited by 112
  32. 2025OneTwoVLA: A Unified Vision-Language-Action Model with Adaptive Reasoningcited by 105
  33. 2018Imitating Latent Policies from Observationcited by 101
  34. 2025UP-VLA: A Unified Understanding and Prediction Model for Embodied Agentcited by 99
  35. 2026A Pragmatic VLA Foundation Modelcited by 97
  36. 2025Discrete Diffusion VLA: Bringing Discrete Diffusion to Action Decoding in Vision-Language-Action Policiescited by 90
  37. 2024What Matters in Building Vision-Language-Action Models for Generalist Robotscited by 88
  38. 2025F1: A Vision-Language-Action Model Bridging Understanding and Generation to Actionscited by 85

50 most cited of 71 All 71 (JSON)

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2026: 104.

104 papers.

Sort:

By area:

  1. 2026Being-H0.7: A Latent World-Action Model from Egocentric Videoscited by 68
  2. 2026Unified 4D World Action Modeling from Video Priors with Asynchronous Denoisingcited by 65
  3. 2026World Action Models: The Next Frontier in Embodied AIcited by 46
  4. 2026LaWAM: Latent World Action Models for Efficient Dynamics-Aware Robot Policiescited by 46
  5. 2026World Model for Robot Learning: A Comprehensive Surveycited by 42
  6. 2026Do World Action Models Generalize Better than VLAs? A Robustness Studycited by 35
  7. 2026ImageWAM: Do World Action Models Really Need Video Generation, or Just Image Editing?cited by 33
  8. 2026ABot-M0.5: Unified Mobility-and-Manipulation World Action Modelcited by 28
  9. 2026OA-WAM: Object-Addressable World Action Model for Robust Robot Manipulationcited by 25
  10. 2026MaskWAM: Unifying Mask Prompting and Prediction for World-Action Modelscited by 15
  11. 2026JEPA-WAM: Learning Vision-Language-Action Policies with Joint-Embedding World Modelingcited by 12
  12. 2026World Action Models: A Surveycited by 10
38 more
  1. 2026ST-WAM: Semantic-Temporal World Action Model for Robust Manipulation under Visual Distribution Shiftscited by 10
  2. 2026Feedback World Model Enables Precise Guidance of Diffusion Policycited by 9
  3. 2026TurboVLA: Real-Time Vision-Language-Action Model at 32 Hz on an RTX 4090 with <1 GB VRAMcited by 9
  4. 2026Faster-WAM: Do World Action Models Need Deep Action Modules?cited by 9
  5. 2026LIBERO-Safety: A Comprehensive Benchmark for Physical and Semantic Safety in Vision-Language-Action Modelscited by 8
  6. 2026IntentVLA: Short-Horizon Intent Modeling for Aliased Robot Manipulationcited by 7
  7. 2026FrameSkip: Learning from Fewer but More Informative Frames in VLA Trainingcited by 6
  8. 2026Orca: The World is in Your Mindcited by 6
  9. 2026DreamWAM: Beyond RGB Future Prediction for World Action Modelscited by 6
  10. 2026Human Cognition in Machines: A Unified Perspective of World Modelscited by 5
  11. 2026AffordVLA: Injecting Affordance Representations into Vision-Language-Action Models via Implicit Feature Alignmentcited by 5
  12. 2026APT: Action Expert Pretraining Improves Instruction Generalization of Vision-Language-Action Policiescited by 5
  13. 2026ACE-Brain-0.5: A Unified Embodied Foundational Model for Physical Agentic AIcited by 5
  14. 2026Track4Action: Distilling World-Centric 3D Tracker into Vision-Language-Action Policiescited by 5
  15. 2026vla.cpp: A Unified Inference Runtime for Vision-Language-Action Modelscited by 4
  16. 2026Latent World Models with Monotone Planning Costs for Image-Goal Navigationcited by 4
  17. 2026Policy-Guided World Model Planning for Language-Conditioned Visual Navigationcited by 3
  18. 2026CKT-WAM: Parameter-Efficient Context Knowledge Transfer Between World Action Modelscited by 3
  19. 2026ALAM: Algebraically Consistent Latent Action Model for Vision-Language-Action Modelscited by 3
  20. 2026A Definition and Roadmap for World Modelscited by 3
  21. 2026Jetson-PI: Towards Onboard Real-Time Robot Control via Foresight-Aligned Asynchronous Inferencecited by 3
  22. 2026Enfold: Folding World Model Imagination into Predictive Representations for Ultra-Efficient Embodied Controlcited by 3
  23. 2026SG-WAM: Self-Guided World Modeling in Geometry-Aware Policy Spacecited by 3
  24. 2026StageWAM: Joint-Embedding Stage Prediction for World-Action Models in Robot Manipulationcited by 3
  25. 2026From Imagined Futures to Executable Actions: Mixture of Latent Actions for Robot Manipulationcited by 2
  26. 2026World-Ego Modeling for Long-Horizon Evolution in Hybrid Embodied Taskscited by 2
  27. 2026QuoVLA: Quotient Space for Vision-Language-Action Modelscited by 2
  28. 2026ω-EVA: Envision, Verify, and Act with Latent Interactive World Modelscited by 2
  29. 2026VLAFlow: A Unified Training Framework for Vision-Language-Action Models via Co-training and Future Latent Alignmentcited by 2
  30. 2026Towards Predictive, Aligned, and Scalable Robot Learningcited by 2
  31. 2026FoMoVLA: Bridging Visual Foresight and Motion Guidance for Vision-Language-Action Modelscited by 2
  32. 2026CheckVLA: Execution-Time Verification with Action-Conditioned World Model for Long-Horizon Mobile Manipulationcited by 2
  33. 2026The Geometry of Flow-Matching Uncertainty: A Cost-free Uncertainty Proxy and Its Application in Flow-based VLA Failure Detectioncited by 2
  34. 2026EgoGenesis: Egocentric World-Action Modeling with Online Anchored Projective Memory and Action-3D RoPEcited by 2
  35. 2026GWM-VLA: Geometry-Aware Latent World Modeling for Vision-Language-Action Learningcited by 2
  36. 2026Learning Action Priors for Cross-embodiment Robot Manipulationcited by 1
  37. 2026From World Action Models to Embodied Brains: A Roadmap for Open-World Physical Intelligencecited by 1
  38. 2026Anticipate Before Acting: Future-State-Conditioned Vision-Language Navigationcited by 1

50 most cited of 104 All 104 (JSON)

Similar, not cited

Close in wording, no citation either way

  1. 2026JEPA-VLA: Video Predictive Embedding is Needed for VLA Models
  2. 2026ViDAL: A Visual Dynamics-Grounded Action Latent Space for Vision-Language-Action Models
  3. 2026From Pixels to Tokens: A Systematic Study of Latent Action Supervision for Vision-Language-Action Models
  4. 2026World2Act: Latent Action Post-Training from World Model Dynamics
  5. 2026One Token Per Frame: Reconsidering Visual Bandwidth in World Models for VLA Policy
  6. 2026LARA: Latent Action Representation Alignment for Vision-Language-Action Models
  7. 2026Chain of World: World Model Thinking in Latent Motion
  8. 2026Beyond World-Frame Action Heads: Motion-Centric Action Frames for Vision-Language-Action Models