mlsanityData to 7 Oct 2026

Motus: A Unified Latent Action World Model

arXiv 2512.13030 ↗First posted 15 Dec 2025cs.CV cs.LG cs.ROcited by 231

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

Abstract (arXiv)

While a general embodied agent must function as a unified system, current methods are built on isolated models for understanding, world modeling, and control. This fragmentation prevents unifying multimodal generative capabilities and hinders learning from large-scale, heterogeneous data. In this paper, we propose Motus, a unified latent action world model that leverages existing general pretrained models and rich, sharable motion information. Motus introduces a Mixture-of-Transformer (MoT) architecture to integrate three experts (i.e., understanding, video generation, and action) and adopts a UniDiffuser-style scheduler to enable flexible switching between different modeling modes (i.e., world models, vision-language-action models, inverse dynamics models, video generation models, and video-action joint prediction models). Motus further leverages the optical flow to learn latent actions and adopts a recipe with three-phase training pipeline and six-layer data pyramid, thereby extracting pixel-level "delta action" and enabling large-scale action pretraining. Experiments show that Motus achieves superior performance against state-of-the-art methods in both simulation (a +15% improvement over X-VLA and a +45% improvement over Pi0.5) and real-world scenarios(improved by +11~48%), demonstrating unified modeling of all functionalities and priors significantly benefits downstream robotic tasks.

Key facts. Cited by 231 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 58 of its 62 references that are in the map. First posted on arXiv 15 Dec 2025.

Builds on

58 of 62 references.

Closest ancestors · 9

Same direction, newest first

  1. 2025FlowVLA: Visual Chain of Thought-based Motion Reasoning for Vision-Language-Action Models25 Aug
  2. 2025Latent Policy Steering with Embodiment-Agnostic Pretrained World Models17 Jul
  3. 2025AMPLIFY: Actionless Motion Priors for Robot Learning from Videos17 Jun
  4. 2025Unified World Models: Coupling Video and Action Diffusion for Pretraining on Large Robotic Datasets3 Apr
  5. 2025AdaWorld: Learning Adaptable World Models with Latent Actions24 Mar
  6. 2024Predictive Inverse Dynamics Models are Scalable Learners for Robotic Manipulation19 Dec
  7. 2024RoboDreamer: Learning Compositional World Models for Robot Imagination18 Apr
  8. 2023Learning to Act without Actions17 Dec
  9. 2023Zero-Shot Robotic Manipulation with Pretrained Image-Editing Diffusion Models16 Oct

Other references · 49

  1. 2025Qwen2.5-VL Technical Reportcited by 5,418
  2. 2024Qwen2-VL: Enhancing Vision-Language Model's Perception of the World at Any Resolutioncited by 4,151
  3. 2023RT-2: Vision-Language-Action Models Transfer Web Knowledge to Robotic Controlcited by 2,993
  4. 2024OpenVLA: An Open-Source Vision-Language-Action Modelcited by 2,777
  5. 2023Qwen-VL: A Versatile Vision-Language Model for Understanding, Localization, Text Reading, and Beyondcited by 2,179
  6. 2025π0.5: a Vision-Language-Action Model with Open-World Generalizationcited by 1,653
  7. 2023Learning Fine-Grained Bimanual Manipulation with Low-Cost Hardwarecited by 1,490
  8. 2025GR00T N1: An Open Foundation Model for Generalist Humanoid Robotscited by 1,167
  9. 2024Chameleon: Mixed-Modal Early-Fusion Foundation Modelscited by 904
  10. 2025Emerging Properties in Unified Multimodal Pretrainingcited by 855
  11. 2024Emu3: Next-Token Prediction is All You Needcited by 761
  12. 2024Genie: Generative Interactive Environmentscited by 706
37 more
  1. 2024RDT-1B: a Diffusion Foundation Model for Bimanual Manipulationcited by 670
  2. 2023Learning Universal Policies via Text-Guided Video Generationcited by 506
  3. 2025RoboTwin 2.0: A Scalable Data Generator and Benchmark with Strong Domain Randomization for Robust Bimanual Robotic Manipulationcited by 477
  4. 2024Janus: Decoupling Visual Encoding for Unified Multimodal Understanding and Generationcited by 447
  5. 2023Learning Interactive Real-World Simulatorscited by 440
  6. 2025AgiBot World Colosseo: A Large-scale Manipulation Platform for Scalable and Intelligent Embodied Systemscited by 390
  7. 2025UniVLA: Learning to Act Anywhere with Task-centric Latent Actionscited by 386
  8. 2024Latent Action Pretraining from Videoscited by 302
  9. 2025X-VLA: Soft-Prompted Transformer as Scalable Cross-Embodiment Vision-Language-Action Modelcited by 285
  10. 2024Video Prediction Policy: A Generalist Robot Policy with Predictive Visual Representationscited by 279
  11. 2025Show-o2: Improved Native Unified Multimodal Modelscited by 242
  12. 2025MMaDA: Multimodal Large Diffusion Language Modelscited by 241
  13. 2024Deep Compression Autoencoder for Efficient High-Resolution Diffusion Modelscited by 209
  14. 2025Unified Video Action Modelcited by 195
  15. 2025EgoDex: Learning Dexterous Manipulation from Large-Scale Egocentric Videocited by 191
  16. 2024RoboMIND: Benchmark on Multi-embodiment Intelligence Normative Data for Robot Manipulationcited by 166
  17. 2024Mixture-of-Transformers: A Sparse and Scalable Architecture for Multi-Modal Foundation Modelscited by 151
  18. 2024Scaling Proprioceptive-Visual Learning with Heterogeneous Pre-trained Transformerscited by 145
  19. 2024Gen2Act: Human Video Generation in Novel Scenarios enables Generalizable Robot Manipulationcited by 142
  20. 2025VideoJAM: Joint Appearance-Motion Representations for Enhanced Motion Generation in Video Modelscited by 112
  21. 2018Imitating Latent Policies from Observationcited by 101
  22. 2025F1: A Vision-Language-Action Model Bridging Understanding and Generation to Actionscited by 85
  23. 2024Dual Diffusion for Unified Image Generation and Understandingcited by 76
  24. 2024Moto: Latent Motion Token as the Bridging Language for Learning Robot Manipulation from Videoscited by 65
  25. 2025Vidar: Embodied Video Diffusion Model for Generalist Manipulationcited by 58
  26. 2025HERMES: A Unified Self-Driving World Model for Simultaneous 3D Scene Understanding and Generationcited by 56
  27. 2025Latent Action Learning Requires Supervision in the Presence of Distractorscited by 40
  28. 2025CoMo: Learning Continuous Latent Motion from Internet Videos for Scalable Robot Learningcited by 37
  29. 2025ShapeLLM-Omni: A Native Multimodal LLM for 3D Generation and Understandingcited by 34
  30. 2025DPFlow: Adaptive Optical Flow Estimation with a Dual-Pyramid Frameworkcited by 30
  31. 2025What Do Latent Action Models Actually Learn?cited by 30
  32. 2018Learning what you can do before doing anythingcited by 26
  33. 2025Action-Free Reasoning for Policy Generalizationcited by 24
  34. 2025AnyPos: Automated Task-Agnostic Actions for Bimanual Manipulationcited by 23
  35. 2025MotionTrans: Human VR Data Enable Motion-Level Learning for Robotic Manipulation Policiescited by 22
  36. 2025Video2Policy: Scaling up Manipulation Tasks in Simulation through Internet Videoscited by 21
  37. 2024Tra-MoE: Learning Trajectory Prediction Model from Multiple Domains for Adaptive Policy Conditioningcited by 18

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2025: 1; 2026: 230.

231 papers.

Sort:

By area:

  1. 2026Causal World Modeling for Robot Controlcited by 289
  2. 2026Cosmos 3: Omnimodal World Models for Physical AIcited by 137
  3. 2026GigaWorld-Policy: An Efficient Action-Centered World--Action Modelcited by 129
  4. 2026VLA-JEPA: Enhancing Vision-Language-Action Model with Latent World Modelcited by 104
  5. 2026ABot-M0: VLA Foundation Model for Robotic Manipulation with Action Manifold Learningcited by 94
  6. 2026DiT4DiT: Jointly Modeling Video Dynamics and Actions for Generalizable Robot Controlcited by 71
  7. 2026Being-H0.7: A Latent World-Action Model from Egocentric Videoscited by 68
  8. 2026Unified 4D World Action Modeling from Video Priors with Asynchronous Denoisingcited by 65
  9. 2026Motubrain: An Advanced World Action Model for Robot Controlcited by 54
  10. 2026InternVLA-A1: Unifying Understanding, Generation and Action for Robotic Manipulationcited by 48
  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. 2026RMBench: Memory-Dependent Robotic Manipulation Benchmark with Insights into Policy Designcited by 40
  4. 2026Do World Action Models Generalize Better than VLAs? A Robustness Studycited by 35
  5. 2026RoboDojo: A Unified Sim-and-Real Benchmark for Comprehensive Evaluation of Generalist Robot Manipulation Policiescited by 35
  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. 2025VLA-Arena: An Open-Source Framework for Benchmarking Vision-Language-Action Modelscited by 28
  9. 2026ReMem-VLA: Empowering Vision-Language-Action Model with Memory via Dual-Level Recurrent Queriescited by 28
  10. 2026ABot-M0.5: Unified Mobility-and-Manipulation World Action Modelcited by 28
  11. 2026GigaBrain-0.5M*: a VLA That Learns From World Model-Based Reinforcement Learningcited by 23
  12. 2026AtomVLA: Scalable Post-Training for Robotic Manipulation via Predictive Latent World Modelscited by 21
  13. 2026τ0-WM: A Unified Video-Action World Model for Robotic Manipulationcited by 21
  14. 2026HoloBrain-0 Technical Reportcited by 20
  15. 2026Hy-Embodied-0.5-VLA: From Vision-Language-Action Models to a Real-World Robot Learning Stackcited by 20
  16. 2026Light-WAM: Efficient World Action Models with State-Fusion Action Decodingcited by 19
  17. 2026When to Trust Imagination: Adaptive Action Execution for World Action Modelscited by 18
  18. 2026UniDriveVLA: Unifying Understanding, Perception, and Action Planning for Autonomous Drivingcited by 17
  19. 2026Efficient-WAM: A 1B-Parameter World-Action Model with Low-Cost Future Imaginationcited by 17
  20. 2026G0.5: One Autoregressive Stream for Robot Reasoning and Actioncited by 17
  21. 2026Flash-WAM: Modality-Aware Distillation for World Action Modelscited by 16
  22. 2026EVA: Aligning Video World Models with Executable Robot Actions via Inverse Dynamics Rewardscited by 14
  23. 2026WALL-WM: Carving World Action Modeling at the Event Jointscited by 14
  24. 2026Next Forcing: Causal World Modeling with Multi-Chunk Predictioncited by 14
  25. 2026Faster-WAM: Efficient Inference-Time Future Conditioning for Robust World Action Modelscited by 14
  26. 2026HALO: A Unified Vision-Language-Action Model for Embodied Multimodal Chain-of-Thought Reasoningcited by 13
  27. 2026World Action Verifier: Self-Improving World Models via Forward-Inverse Asymmetrycited by 13
  28. 2026World-Value-Action Model: Implicit Planning for Vision-Language-Action Systemscited by 13
  29. 2026VLANeXt: Recipes for Building Strong VLA Modelscited by 12
  30. 2026AIM: Intent-Aware Unified world action Modeling with Spatial Value Mapscited by 12
  31. 2026WAM4D: Fast 4D World Action Model via Spatial Register Tokenscited by 12
  32. 2026MemoryWAM: Efficient World Action Modeling with Persistent Memorycited by 12
  33. 2026JEPA-WAM: Learning Vision-Language-Action Policies with Joint-Embedding World Modelingcited by 12
  34. 2026STARRY: Spatial-Temporal Action-Centric World Modeling for Robotic Manipulationcited by 10
  35. 2026Is the Future Compatible? Diagnosing Dynamic Consistency in World Action Modelscited by 10
  36. 2026HarmoWAM: Harmonizing Generalizable and Precise Manipulation via Adaptive World Action Modelscited by 10
  37. 2026World-Language-Action Model for Unified World Modeling, Language Reasoning, and Action Synthesiscited by 10
  38. 2026Metis: A Generalizable and Efficient World-Action Model for Autonomous Driving and Urban Navigationcited by 10

50 most cited of 231 All 231 (JSON)

Similar, not cited

Close in wording, no citation either way

  1. 2025MoWM: Mixture-of-World-Models for Embodied Planning via Latent-to-Pixel Feature Modulation
  2. 2026LeapBot-WA: World-Anchor Action Models via Predictive Latent Alignments
  3. 2026OpenWAM: An Open, Modular Exploration Towards Systematic World-Action Model Pretraining
  4. 2026Learning Latent Action World Models In The Wild
  5. 2026Worldscape-MoE: A Unified Mixture-of-Experts World Model for Scalable Heterogeneous Action Control
  6. 2026Chain of World: World Model Thinking in Latent Motion
  7. 2026LaWM: Least Action World Models for Long-Horizon Physical Consistency from Visual Observations
  8. 2026UniWAM: Unified World-Action Model