mlsanityData to 7 Oct 2026

World model-based VLA

76 papers3 this weekHigh confidence in this grouping: the papers form one clear topicHigh confidence in this grouping: the papers form one clear topic · About

Keywords: action, visual, modeling, latent, vla

Key facts. World model-based VLA: 76 papers on arXiv since 2025 (data through 7 Oct 2026). 22 of its 76 mapped papers (29%) cite “DreamVLA: A Vision-Language-Action Model Dreamed with Comprehensive World Knowledge” (2025, arXiv 2507.04447). 15 of its 76 mapped papers (20%) cite “World Action Models are Zero-shot Policies” (2026, arXiv 2602.15922). 18 new papers 17 Sep–7 Oct.

What’s new

17 Sep – 7 Oct 2026

18 new papers, 17 Sep – 7 Oct; 3 this week.

  1. This week · 3
  2. Video Prediction Policy 2: Predict Better, Act Better also fits: Human-video and latent-action pretraining
  3. Juno: Taming Predictive Latents for Vision-Language-Action Models also fits: VLA architectures and generalization
  4. ViDAL: A Visual Dynamics-Grounded Action Latent Space for Vision-Language-Action Models also fits: Human-video and latent-action pretraining
  5. Earlier · 15
  6. ForeAct3D: Policy-Grounded Future World Modeling for VLA Policies
  7. SUAVE: Unified Video-Action Models via Masked Diffusion also fits: Efficient VLA inference and adaptation
  8. Imagine the Future, Internalize the Gist: Efficient VLA Reasoning via Internalized Spatiotemporal Imagination also fits: Reasoning and chain-of-thought in VLA
  9. World-Calibrated Proposal-to-Action Flow for Vision-Language-Action Models
  10. UniWAM: Unified World-Action Model
  11. ATI-VLA: Action-Centric Predictive Vision-Language-Action Models via Actionable Alignment Then Adaptive Injection also fits: Efficient VLA inference and adaptation
  12. Devol-ONE: One Autoregressive Mixture of Transformers to Unify Vision-Language-Action and Latent World Modeling
All 18
  1. FRAM: Trajectory-Guided Visual Feature Selection for Compact Language-Conditioned Robot Manipulation also fits: Human-video and latent-action pretraining
  2. Token-World: World Modeling in Vision-Language Model Token Space for Robot Manipulation
  3. EWAM: Emergent Depth-Wise Specialization in a Unified Embodied Model -- From Semantic Understanding through Visual Foresight to Action
  4. MotionWeave: Learning Motion-Centered Future Dynamics for Vision-Language-Action Policies
  5. WorldGuide: Learning Success-Failure Boundaries in Latent World Models for Vision-Language-Action Policies also fits: Reinforcement learning and post-training
  6. SLIP-VLA: Single-Step Latent Imagination for Policy Learning in Vision-Language-Action Models
  7. Towards VLA-Dreamer: Refining VLA Behavior Using World Models also fits: Generalist VLA models and benchmarks
  8. Think Like a World Model, Act Like a VLA: Distilling World-Model Representations into Compact Robot Policies

Foundations

Most cited by papers here

  1. DreamVLA: A Vision-Language-Action Model Dreamed with Comprehensive World Knowledge2025cited by 22 here
  2. World Action Models are Zero-shot Policies2026cited by 15 here
  3. Fast-WAM: Do World Action Models Need Test-time Future Imagination?2026cited by 14 here
  4. VLA-JEPA: Enhancing Vision-Language-Action Model with Latent World Model2026cited by 12 here
  5. F1: A Vision-Language-Action Model Bridging Understanding and Generation to Actions2025cited by 11 here

Lineage

The papers this direction builds on, oldest first.

Each paper on the line cites the one before.

  1. 2023RT-2: Vision-Language-Action Models Transfer Web Knowledge to Robotic Controlcited by 46 here
  2. 2024Octo: An Open-Source Generalist Robot Policycited by 38 here
  3. 2024OpenVLA: An Open-Source Vision-Language-Action Modelcited by 44 here
  4. 2024π0: A Vision-Language-Action Flow Model for General Robot Controlcited by 44 here
  5. 2025FAST: Efficient Action Tokenization for Vision-Language-Action Modelscited by 19 here
  6. 2025π0.5: a Vision-Language-Action Model with Open-World Generalizationcited by 37 here
  7. 2025DreamVLA: A Vision-Language-Action Model Dreamed with Comprehensive World Knowledgecited by 22 here
  8. 2026Fast-WAM: Do World Action Models Need Test-time Future Imagination?cited by 14 here
16 more it builds on, not in this chain
  1. 20243D-VLA: A 3D Vision-Language-Action Generative World Modelcited by 9 here
  2. 2024CogACT: A Foundational Vision-Language-Action Model for Synergizing Cognition and Action in Robotic Manipulationcited by 15 here
  3. 2024TraceVLA: Visual Trace Prompting Enhances Spatial-Temporal Awareness for Generalist Robotic Policiescited by 11 here
  4. 2024Video Prediction Policy: A Generalist Robot Policy with Predictive Visual Representationscited by 22 here
  5. 2025SpatialVLA: Exploring Spatial Representations for Visual-Language-Action Modelcited by 16 here
  6. 2025UP-VLA: A Unified Understanding and Prediction Model for Embodied Agentcited by 12 here
  7. 2025Fine-Tuning Vision-Language-Action Models: Optimizing Speed and Successcited by 26 here
  8. 2025CoT-VLA: Visual Chain-of-Thought Reasoning for Vision-Language-Action Modelscited by 25 here
  9. 2025UniVLA: Learning to Act Anywhere with Task-centric Latent Actionscited by 24 here
  10. 2025Unified Vision-Language-Action Modelcited by 13 here
  11. 2025F1: A Vision-Language-Action Model Bridging Understanding and Generation to Actionscited by 11 here
  12. 2025X-VLA: Soft-Prompted Transformer as Scalable Cross-Embodiment Vision-Language-Action Modelcited by 13 here
  13. 2025LIBERO-Plus: In-depth Robustness Analysis of Vision-Language-Action Modelscited by 14 here
  14. 2025RynnVLA-002: A Unified Vision-Language-Action and World Modelcited by 10 here
  15. 2026VLA-JEPA: Enhancing Vision-Language-Action Model with Latent World Modelcited by 12 here
  16. 2026World Action Models are Zero-shot Policiescited by 15 here

The 24 most cited of 256.

Browse

30 of 76, ranked 6–35

  1. 6RynnVLA-002: A Unified Vision-Language-Action and World Model202510
  2. 7Ctrl-World: A Controllable Generative World Model for Robot Manipulation20258
  3. 8World Guidance: World Modeling in Condition Space for Action Generation20265
  4. 9JEPA-VLA: Video Predictive Embedding is Needed for VLA Models20264
  5. 10OA-WAM: Object-Addressable World Action Model for Robust Robot Manipulation20264
  6. 11FutureVLA: Joint Visuomotor Prediction for Vision-Language-Action Model20263
  7. 12World-Value-Action Model: Implicit Planning for Vision-Language-Action Systems20263
  8. 13Motubrain: An Advanced World Action Model for Robot Control20263
  9. 14World Action Models: The Next Frontier in Embodied AI20263
  10. 15Pixel Motion as Universal Representation for Robot Control20253
All 30
  1. 16FRAPPE: Infusing World Modeling into Generalist Policies via Multiple Future Representation Alignment20262
  2. 17Global Prior Meets Local Consistency: Dual-Memory Augmented Vision-Language-Action Model for Efficient Robotic Manipulation20262
  3. 18Chain of World: World Model Thinking in Latent Motion20262
  4. 19DeltaVLA: Prior-Guided Vision-Language-Action Models via World Knowledge Variation20262
  5. 20Do World Action Models Generalize Better than VLAs? A Robustness Study20262
  6. 21MinD: Learning A Dual-System World Model for Real-Time Planning and Implicit Risk Analysis20252
  7. 22Mantis: A Versatile Vision-Language-Action Model with Disentangled Visual Foresight20252
  8. 23Unifying Perception and Action: A Hybrid-Modality Pipeline with Implicit Visual Chain-of-Thought for Robotic Action Generation20252
  9. 24ForeAct: Steering Your VLA with Efficient Visual Foresight Planning20261
  10. 25World2Act: Latent Action Post-Training from World Model Dynamics20261
  11. 26DIAL: Decoupling Intent and Action via Latent World Modeling for End-to-End VLA20261
  12. 27DriveDreamer-Policy: A Geometry-Grounded World-Action Model for Unified Generation and Planning20261
  13. 28Mask World Model: Predicting What Matters for Robust Robot Policy Learning20261
  14. 29Test-Time Training for Visual Foresight Vision-Language-Action Models20261
  15. 30Intercepting the Future: Latent-Space Predictive World Model for Dynamic VLA Manipulation20261
  16. 31LaWAM: Latent World Action Models for Efficient Dynamics-Aware Robot Policies20261
  17. 32DSWAM: A Dual-System World Action Foundation Model for Fine-Grained Robot Manipulation20261
  18. 33JEPA-WAM: Learning Vision-Language-Action Policies with Joint-Embedding World Modeling20261
  19. 34StageWAM: Joint-Embedding Stage Prediction for World-Action Models in Robot Manipulation20261
  20. 35TriVLA: A Triple-System-Based Unified Vision-Language-Action Model with Episodic World Modeling for General Robot Control20251