mlsanityData to 7 Oct 2026

FlowVLA: Visual Chain of Thought-based Motion Reasoning for Vision-Language-Action Models

arXiv 2508.18269 ↗First posted 25 Aug 2025cs.ROcited by 50

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

also in 2 more
Abstract (arXiv)

Many Vision-Language-Action (VLA) models are built upon an internal world model trained via next-frame prediction ``vt rightarrow vt+1''. However, this paradigm attempts to predict the future frame's appearance directly, without explicitly reasoning about the underlying dynamics. This lack of an explicit motion reasoning step often leads to physically implausible visual forecasts and inefficient policy learning. To address this limitation, we introduce the Visual Chain of Thought (Visual CoT), a paradigm that compels the model to first reason about motion dynamics before generating the future frame. We instantiate this paradigm by proposing FlowVLA, an autoregressive Transformer that explicitly materializes this reasoning process as ``vt rightarrow ft rightarrow vt+1'', where ft is an intermediate optical flow prediction that inherently encodes motion. By forcing the model to first follow the motion plan encoded by ft, this process inherently aligns the pre-training objective of dynamics prediction with the downstream task of action generation. We conduct experiments on challenging robotics manipulation benchmarks, as well as real-robot evaluations. Our FlowVLA not only generates more coherent and physically plausible visual predictions, but also achieves state-of-the-art policy performance with substantially improved sample efficiency, pointing toward a more principled foundation for world modeling in VLAs. Project page: https://irpn-lab.github.io/FlowVLA/

Key facts. Cited by 50 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 33 of its 37 references that are in the map. First posted on arXiv 25 Aug 2025.

Builds on

33 of 37 references.

Closest ancestors · 3

Same direction, newest first

  1. 2025WorldVLA: Towards Autoregressive Action World Model26 Jun
  2. 2024Robotic Control via Embodied Chain-of-Thought Reasoning11 Jul
  3. 2023Diffusion Policy: Visuomotor Policy Learning via Action Diffusion7 Mar

Other references · 30

  1. 2022Chain-of-Thought Prompting Elicits Reasoning in Large Language Modelscited by 12,302
  2. 2020Taming Transformers for High-Resolution Image Synthesiscited by 3,229
  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. 2020RAFT: Recurrent All-Pairs Field Transforms for Optical Flowcited by 2,385
  6. 2023Learning Fine-Grained Bimanual Manipulation with Low-Cost Hardwarecited by 1,490
  7. 2018World Modelscited by 1,444
  8. 2024Octo: An Open-Source Generalist Robot Policycited by 1,427
  9. 2023LIBERO: Benchmarking Knowledge Transfer for Lifelong Robot Learningcited by 1,338
  10. 2025Fine-Tuning Vision-Language-Action Models: Optimizing Speed and Successcited by 844
  11. 2025FAST: Efficient Action Tokenization for Vision-Language-Action Modelscited by 643
  12. 2025SpatialVLA: Exploring Spatial Representations for Visual-Language-Action Modelcited by 502
18 more
  1. 2024Evaluating Real-World Robot Manipulation Policies in Simulationcited by 484
  2. 2025CoT-VLA: Visual Chain-of-Thought Reasoning for Vision-Language-Action Modelscited by 440
  3. 2024Mobile ALOHA: Learning Bimanual Mobile Manipulation with Low-Cost Whole-Body Teleoperationcited by 377
  4. 2023Unleashing Large-Scale Video Generative Pre-training for Visual Robot Manipulationcited by 353
  5. 2024TraceVLA: Visual Trace Prompting Enhances Spatial-Temporal Awareness for Generalist Robotic Policiescited by 298
  6. 2024RoboPoint: A Vision-Language Model for Spatial Affordance Prediction for Roboticscited by 242
  7. 2025ThinkAct: Vision-Language-Action Reasoning via Reinforced Visual Latent Planningcited by 152
  8. 2025Unified Vision-Language-Action Modelcited by 128
  9. 2025VideoJAM: Joint Appearance-Motion Representations for Enhanced Motion Generation in Video Modelscited by 112
  10. 2024What Matters in Building Vision-Language-Action Models for Generalist Robotscited by 88
  11. 2025PD-VLA: Accelerating Vision-Language-Action Model Integrated with Action Chunking via Parallel Decodingcited by 88
  12. 2025ReconVLA: Reconstructive Vision-Language-Action Model as Effective Robot Perceivercited by 80
  13. 2025Dita: Scaling Diffusion Transformer for Generalist Vision-Language-Action Policycited by 73
  14. 2025CEED-VLA: Consistency Vision-Language-Action Model with Early-Exit Decodingcited by 38
  15. 2024Learning Manipulation by Predicting Interactioncited by 36
  16. 2025Embodied-R1: Reinforced Embodied Reasoning for General Robotic Manipulationcited by 36
  17. 2025From Seeing to Doing: Bridging Reasoning and Decision for Robotic Manipulationcited by 34
  18. 2022Goal-conditioned dual-action imitation learning for dexterous dual-arm robot manipulationcited by 21

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2025: 9; 2026: 41.

50 papers.

Sort:

By area:

  1. 2025Motus: A Unified Latent Action World Modelcited by 231
  2. 2025VLA-Adapter: An Effective Paradigm for Tiny-Scale Vision-Language-Action Modelcited by 140
  3. 2026VLA-JEPA: Enhancing Vision-Language-Action Model with Latent World Modelcited by 104
  4. 2025F1: A Vision-Language-Action Model Bridging Understanding and Generation to Actionscited by 85
  5. 2025VLA-RFT: Vision-Language-Action Reinforcement Fine-tuning with Verified Rewards in World Simulatorscited by 65
  6. 2025Dual-Stream Diffusion for World-Model Augmented Vision-Language-Action Modelcited by 33
  7. 2026When Vision Overrides Language: Evaluating and Mitigating Counterfactual Failures in VLAscited by 27
  8. 2026JEPA-VLA: Video Predictive Embedding is Needed for VLA Modelscited by 22
  9. 2025Mixture of Horizons in Action Chunkingcited by 20
  10. 2026DualCoT-VLA: Visual-Linguistic Chain of Thought via Parallel Reasoning for Vision-Language-Action Modelscited by 20
  11. 2025AstraNav-World: World Model for Foresight Control and Consistencycited by 17
  12. 2026Recurrent-Depth VLA: Implicit Test-Time Compute Scaling of Vision-Language-Action Models via Latent Iterative Reasoningcited by 17
38 more
  1. 2026VP-VLA: Visual Prompting as an Interface for Vision-Language-Action Modelscited by 17
  2. 2026G0.5: One Autoregressive Stream for Robot Reasoning and Actioncited by 17
  3. 2025AsyncVLA: Asynchronous Flow Matching for Vision-Language-Action Modelscited by 16
  4. 2026World-Value-Action Model: Implicit Planning for Vision-Language-Action Systemscited by 13
  5. 2026VLANeXt: Recipes for Building Strong VLA Modelscited by 12
  6. 2026LaMP: Learning Vision-Language-Action Policy with 3D Scene Flow as Latent Motion Priorcited by 11
  7. 2026SimVLA: A Simple VLA Baseline for Robotic Manipulationcited by 10
  8. 2026FlowWAM: Optical Flow as a Unified Action Representation for World Action Modelscited by 10
  9. 2026Chain of World: World Model Thinking in Latent Motioncited by 8
  10. 2026AffordanceVLA: A Vision-Language-Action Model Empowering Action Generation through Affordance-Aware Understandingcited by 8
  11. 2025Robotic VLA Benefits from Joint Learning with Motion Image Diffusioncited by 7
  12. 2026ReViP: Mitigating False Completion in Vision-Language-Action Models with Vision-Proprioception Rebalancecited by 5
  13. 2026FocusVLA: Focused Visual Utilization for Vision-Language-Action Modelscited by 5
  14. 2026Is Forward Prediction Enough? Physical State Grounding for JEPA World Modelscited by 5
  15. 2026Action Draft and Verify: A Self-Verifying Framework for Vision-Language-Action Modelcited by 4
  16. 2026Escaping the Diversity Trap in Robotic Manipulation via Anchor-Centric Adaptationcited by 4
  17. 2026From Noise to Intent: Anchoring Generative VLA Policies with Residual Bridgescited by 3
  18. 2026Back to the Familiar Future: Failure Recovery for VLA Policies via Pre-Imagined Milestone Selectioncited by 3
  19. 2026LabVLA: Grounding Vision-Language-Action Models in Scientific Laboratoriescited by 3
  20. 2026V-Co: A Closer Look at Visual Representation Alignment via Co-Denoisingcited by 2
  21. 2026Robust-WAM: Bridging Generative Pretraining and Semantic Foresight in World-Action Modelscited by 2
  22. 2026SG-WAM: Text-Grounded and Spatial-aware Semantic Guidance for World-Action Modelscited by 2
  23. 2026Shape of Thought: Progressive Object Assembly via Visual Chain-of-Thoughtcited by 1
  24. 2026Beyond Dense Futures: World Models as Structured Planners for Robotic Manipulationcited by 1
  25. 2026Overcoming Dynamics-Blindness: Training-Free Pace-and-Path Correction for VLA Modelscited by 1
  26. 2026S2-VLA: State-Space Guided Vision-Language-Action Models for Long-Horizon Manipulationcited by 1
  27. 2026PAVXploreRL: Physical-Action-Visual World Model Reinforcement Learning with Action Explorationcited by 1
  28. 2026DyPES-VLA: Learning Shared Dynamics Priors and Embodiment-Specific Control for Cross-Embodiment Manipulationcited by 1
  29. 2026M2-VLA: Boosting Vision-Language Models for Generalizable Manipulation via Layer Mixture and Meta-Skillscited by 0
  30. 2026X-Imitator: Spatial-Aware Imitation Learning via Bidirectional Action-Pose Interactioncited by 0
  31. 2026WorldDiT: A Unified Diffusion Architecture for World and Action Modelingcited by 0
  32. 2026SpatioLM: Towards General Physical Spatial Intelligence in Vision-Language Modelscited by 0
  33. 2026GeomVLA: Unifying Scene, Motion, and Action in 3Dcited by 0
  34. 2026Uni-LaDiR: Latent Diffusion Unifies Multimodal Reasoningcited by 0
  35. 2026SCULPT-VLA: Learning Structured Control through Staged Action Groundingcited by 0
  36. 2026Devol-ONE: One Autoregressive Mixture of Transformers to Unify Vision-Language-Action and Latent World Modelingcited by 0
  37. 2026SLIP-VLA: Single-Step Latent Imagination for Policy Learning in Vision-Language-Action Modelscited by 0
  38. 2026DSDyn-VLA: A Dual-Stream Dynamic Manipulation Framework with Motion Perception, Future Awareness, and Realtime Correctioncited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2025FutureSightDrive: Thinking Visually with Spatio-Temporal CoT for Autonomous Driving
  2. 2026VisualThink-VLA: Visual Intermediate Reasoning for Effective and Low-Latency Vision-Language-Action Policies
  3. 2025CoT4AD: A Vision-Language-Action Model with Explicit Chain-of-Thought Reasoning for Autonomous Driving
  4. 2026ThinkingVLA: Interleaved Vision and Language Reasoning for Robotic Manipulation
  5. 2026World-Calibrated Proposal-to-Action Flow for Vision-Language-Action Models
  6. 2026Xiaomi OneVL: One-Step Latent Reasoning and Planning with Vision-Language Explanation
  7. 2025DeepThinkVLA: Enhancing Reasoning Capability of Vision-Language-Action Models
  8. 2026Resolving State-Representation Mismatch: State-Space Visual Reasoning for Open-Loop VLA Planning