mlsanityData to 7 Oct 2026

World Action Models are Zero-shot Policies

arXiv 2602.15922 ↗First posted 17 Feb 2026cs.RO cs.CV cs.LGcited by 411

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

also in 2 more
Abstract (arXiv)

State-of-the-art Vision-Language-Action (VLA) models excel at semantic generalization but struggle to generalize to unseen physical motions in novel environments. We introduce DreamZero, a World Action Model (WAM) built upon a pretrained video diffusion backbone. Unlike VLAs, WAMs learn physical dynamics by predicting future world states and actions, using video as a dense representation of how the world evolves. By jointly modeling video and action, DreamZero learns diverse skills effectively from heterogeneous robot data without relying on repetitive demonstrations. This results in over 2x improvement in generalization to new tasks and environments compared to state-of-the-art VLAs in real robot experiments. Crucially, through model and system optimizations, we enable a 14B autoregressive video diffusion model to perform real-time closed-loop control at 7Hz. Finally, we demonstrate two forms of cross-embodiment transfer: video-only demonstrations from other robots or humans yield a relative improvement of over 42% on unseen task performance with just 10-20 minutes of data. More surprisingly, DreamZero enables few-shot embodiment adaptation, transferring to a new embodiment with only 30 minutes of play data while retaining zero-shot generalization.

Key facts. Cited by 411 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 80 of its 87 references that are in the map. First posted on arXiv 17 Feb 2026.

Builds on

80 of 87 references.

Closest ancestors · 8

Same direction, newest first

  1. 2026Causal World Modeling for Robot Control29 Jan
  2. 2026PointWorld: Scaling 3D World Models for In-The-Wild Robotic Manipulation7 Jan
  3. 2025Dual-Stream Diffusion for World-Model Augmented Vision-Language-Action Model31 Oct
  4. 2025V-JEPA 2: Self-Supervised Video Models Enable Understanding, Prediction and Planning11 Jun
  5. 2025FLARE: Robot Learning with Implicit World Modeling21 May
  6. 2025DreamGen: Unlocking Generalization in Robot Learning through Video World Models19 May
  7. 2025Unified World Models: Coupling Video and Action Diffusion for Pretraining on Large Robotic Datasets3 Apr
  8. 2024RoboDreamer: Learning Compositional World Models for Robot Imagination18 Apr

Other references · 72

  1. 2025Qwen3 Technical Reportcited by 8,195
  2. 2020Scaling Laws for Neural Language Modelscited by 6,411
  3. 2022Classifier-Free Diffusion Guidancecited by 5,717
  4. 2022Flow Matching for Generative Modelingcited by 5,426
  5. 2021On the Opportunities and Risks of Foundation Modelscited by 3,961
  6. 2022Flow Straight and Fast: Learning to Generate and Transfer Data with Rectified Flowcited by 3,550
  7. 2023RT-2: Vision-Language-Action Models Transfer Web Knowledge to Robotic Controlcited by 2,993
  8. 2024OpenVLA: An Open-Source Vision-Language-Action Modelcited by 2,777
  9. 2025Wan: Open and Advanced Large-Scale Video Generative Modelscited by 2,536
  10. 2022Do As I Can, Not As I Say: Grounding Language in Robotic Affordancescited by 2,302
  11. 2024π0: A Vision-Language-Action Flow Model for General Robot Controlcited by 2,297
  12. 2023PaLM-E: An Embodied Multimodal Language Modelcited by 2,037
38 more
  1. 2022RT-1: Robotics Transformer for Real-World Control at Scalecited by 2,007
  2. 2025π0.5: a Vision-Language-Action Model with Open-World Generalizationcited by 1,653
  3. 2019Dream to Control: Learning Behaviors by Latent Imaginationcited by 1,598
  4. 2021Ego4D: Around the World in 3,000 Hours of Egocentric Videocited by 1,530
  5. 2023Mastering Diverse Domains through World Modelscited by 1,193
  6. 2025GR00T N1: An Open Foundation Model for Generalist Humanoid Robotscited by 1,167
  7. 2022Inner Monologue: Embodied Reasoning through Planning with Language Modelscited by 1,057
  8. 2020Mastering Atari with Discrete World Modelscited by 990
  9. 2023Self-Supervised Learning from Images with a Joint-Embedding Predictive Architecturecited by 982
  10. 2023Stochastic Interpolants: A Unifying Framework for Flows and Diffusionscited by 897
  11. 2025Emerging Properties in Unified Multimodal Pretrainingcited by 855
  12. 2024DROID: A Large-Scale In-The-Wild Robot Manipulation Datasetcited by 801
  13. 2023VoxPoser: Composable 3D Value Maps for Robotic Manipulation with Language Modelscited by 733
  14. 2024SpatialVLM: Endowing Vision-Language Models with Spatial Reasoning Capabilitiescited by 679
  15. 2023BridgeData V2: A Dataset for Robot Learning at Scalecited by 634
  16. 2022ProgPrompt: Generating Situated Robot Task Plans using Large Language Modelscited by 598
  17. 2025Self Forcing: Bridging the Train-Test Gap in Autoregressive Video Diffusioncited by 569
  18. 2023Learning Universal Policies via Text-Guided Video Generationcited by 506
  19. 2023UniPC: A Unified Predictor-Corrector Framework for Fast Sampling of Diffusion Modelscited by 441
  20. 2023Learning Interactive Real-World Simulatorscited by 440
  21. 2025CoT-VLA: Visual Chain-of-Thought Reasoning for Vision-Language-Action Modelscited by 440
  22. 2024From Slow Bidirectional to Fast Autoregressive Video Diffusion Modelscited by 401
  23. 2025UniVLA: Learning to Act Anywhere with Task-centric Latent Actionscited by 386
  24. 2024Pyramidal Flow Matching for Efficient Video Generative Modelingcited by 354
  25. 2023Unleashing Large-Scale Video Generative Pre-training for Visual Robot Manipulationcited by 353
  26. 2024Latent Action Pretraining from Videoscited by 302
  27. 2024TraceVLA: Visual Trace Prompting Enhances Spatial-Temporal Awareness for Generalist Robotic Policiescited by 298
  28. 2024GR-2: A Generative Video-Language-Action Model with Web-Scale Knowledge for Robot Manipulationcited by 290
  29. 2024Video Prediction Policy: A Generalist Robot Policy with Predictive Visual Representationscited by 279
  30. 2026Cosmos Policy: Fine-Tuning Video Models for Visuomotor Control and Planningcited by 274
  31. 2025MAGI-1: Autoregressive Video Generation at Scalecited by 262
  32. 2024Timestep Embedding Tells: It's Time to Cache for Video Diffusion Modelcited by 245
  33. 2025Hi Robot: Open-Ended Instruction Following with Hierarchical Vision-Language-Action Modelscited by 216
  34. 2025Unified Video Action Modelcited by 195
  35. 2025EgoDex: Learning Dexterous Manipulation from Large-Scale Egocentric Videocited by 191
  36. 2025MolmoAct: Action Reasoning Models that can Reason in Spacecited by 189
  37. 2025World Simulation with Video Foundation Models for Physical AIcited by 183
  38. 2023Video Language Planningcited by 160

50 most cited of 72 All 72 (JSON)

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2026: 411.

411 papers.

Sort:

By area:

  1. 2026Cosmos 3: Omnimodal World Models for Physical AIcited by 137
  2. 2026GigaWorld-Policy: An Efficient Action-Centered World--Action Modelcited by 129
  3. 2026π0.7: a Steerable Generalist Robotic Foundation Model with Emergent Capabilitiescited by 127
  4. 2026StarVLA: A Lego-like Codebase for Vision-Language-Action Model Developingcited by 103
  5. 2026Matrix-Game 3.0: Real-Time and Streaming Interactive World Model with Long-Horizon Memorycited by 86
  6. 2026Being-H0.7: A Latent World-Action Model from Egocentric Videoscited by 68
  7. 2026Unified 4D World Action Modeling from Video Priors with Asynchronous Denoisingcited by 65
  8. 2026Motubrain: An Advanced World Action Model for Robot Controlcited by 54
  9. 2026Robometer: Scaling General-Purpose Robotic Reward Models via Trajectory Comparisonscited by 51
  10. 2026SANA-WM: Efficient Minute-Scale World Modeling with Hybrid Linear Diffusion Transformercited by 50
  11. 2026World Action Models: The Next Frontier in Embodied AIcited by 46
  12. 2026LaWAM: Latent World Action Models for Efficient Dynamics-Aware Robot Policiescited by 46
38 more
  1. 2026World Model for Robot Learning: A Comprehensive Surveycited by 42
  2. 2026AHA-WAM:Asynchronous Horizon-Adaptive World-Action Modeling with Observation-Guided Context Routingcited by 41
  3. 2026Solaris: Building a Multiplayer Video World Model in Minecraftcited by 38
  4. 2026DSpark: Confidence-Scheduled Speculative Decoding with Semi-Autoregressive Generationcited by 36
  5. 2026Hierarchical Planning with Latent World Modelscited by 34
  6. 2026ImageWAM: Do World Action Models Really Need Video Generation, or Just Image Editing?cited by 33
  7. 2026Native Video-Action Pretraining for Generalizable Robot Controlcited by 30
  8. 2026ABot-M0.5: Unified Mobility-and-Manipulation World Action Modelcited by 28
  9. 2026Interactive World Simulator for Robot Policy Training and Evaluationcited by 27
  10. 2026RLDX-1 Technical Reportcited by 25
  11. 2026OA-WAM: Object-Addressable World Action Model for Robust Robot Manipulationcited by 25
  12. 2026Xiaomi-Robotics-1: Scaling Vision-Language-Action Models with over 100K Hours of Real-World Trajectoriescited by 24
  13. 2026Causal-rCM: A Unified Teacher-Forcing and Self-Forcing Open Recipe for Autoregressive Diffusion Distillation in Streaming Video Generation and Interactive World Modelscited by 23
  14. 2026Qwen-AgentWorld: Language World Models for General Agentscited by 22
  15. 2026τ0-WM: A Unified Video-Action World Model for Robotic Manipulationcited by 21
  16. 2026DriveVA: Video Action Models are Zero-Shot Driverscited by 19
  17. 2026Light-WAM: Efficient World Action Models with State-Fusion Action Decodingcited by 19
  18. 2026When to Trust Imagination: Adaptive Action Execution for World Action Modelscited by 18
  19. 2026Dream-Tac: A Unified Tactile World Action Model for Contact-Rich Robot Manipulationcited by 17
  20. 2026Efficient-WAM: A 1B-Parameter World-Action Model with Low-Cost Future Imaginationcited by 17
  21. 2026EgoWAM: World Action Models Beyond Pixels with In-the-Wild Egocentric Human Datacited by 17
  22. 2026Show-Harness: Just a VLM Agent Can Play Robotscited by 17
  23. 2026S-VAM: Shortcut Video-Action Model by Self-Distilling Geometric and Semantic Foresightcited by 16
  24. 2026Flash-WAM: Modality-Aware Distillation for World Action Modelscited by 16
  25. 2026MaskWAM: Unifying Mask Prompting and Prediction for World-Action Modelscited by 15
  26. 2026EVA: Aligning Video World Models with Executable Robot Actions via Inverse Dynamics Rewardscited by 14
  27. 2026VTAM: Video-Tactile-Action Models for Complex Physical Interaction Beyond VLAscited by 14
  28. 2026WorldVLN: Autoregressive World Action Model for Aerial Vision-Language Navigationcited by 14
  29. 2026WorldKV: Efficient World Memory with World Retrieval and Compressioncited by 14
  30. 2026WALL-WM: Carving World Action Modeling at the Event Jointscited by 14
  31. 2026NVIDIA OmniDreams: Real-Time Generative World Model for Closed-Loop Autonomous Vehicle Simulationcited by 14
  32. 2026Next Forcing: Causal World Modeling with Multi-Chunk Predictioncited by 14
  33. 2026Learning 4D Geometric Priors for Inference-Efficient World Action Modelscited by 14
  34. 2026Faster-WAM: Efficient Inference-Time Future Conditioning for Robust World Action Modelscited by 14
  35. 2026World Action Verifier: Self-Improving World Models via Forward-Inverse Asymmetrycited by 13
  36. 2026World-Value-Action Model: Implicit Planning for Vision-Language-Action Systemscited by 13
  37. 2026RoboTTT: Context Scaling for Robot Policiescited by 13
  38. 2026VLANeXt: Recipes for Building Strong VLA Modelscited by 12

50 most cited of 411 All 411 (JSON)

Similar, not cited

Close in wording, no citation either way

  1. 2026DreamDojo: A Generalist Robot World Model from Large-Scale Human Videos
  2. 2026OpenWAM: An Open Framework for Composable World-Action Models
  3. 2026Say, Dream, and Act: Learning Video World Models for Instruction-Driven Robot Manipulation
  4. 2025VideoVLA: Video Generators Can Be Generalizable Robot Manipulators
  5. 2026Learning Transferable Dynamics Priors from Action to World Modeling
  6. 2026From Human Videos to Robot Manipulation: A Survey on Scalable Vision-Language-Action Learning with Human-Centric Data
  7. 2026XEWorld: Can Action-Conditioned World Models Generalize to Unseen Robot Embodiments?
  8. 2026CLAP: Cross-Embodiment Video World Models are Zero-Shot Physical Simulators