mlsanityData to 7 Oct 2026

FAST: Efficient Action Tokenization for Vision-Language-Action Models

arXiv 2501.09747 ↗First posted 16 Jan 2025cs.RO cs.LGcited by 643

On the map: Vision-language-action models → Efficient VLA inference and adaptation

Abstract (arXiv)

Autoregressive sequence models, such as Transformer-based vision-language action (VLA) policies, can be tremendously effective for capturing complex and generalizable robotic behaviors. However, such models require us to choose a tokenization of our continuous action signals, which determines how the discrete symbols predicted by the model map to continuous robot actions. We find that current approaches for robot action tokenization, based on simple per-dimension, per-timestep binning schemes, typically perform poorly when learning dexterous skills from high-frequency robot data. To address this challenge, we propose a new compression-based tokenization scheme for robot actions, based on the discrete cosine transform. Our tokenization approach, Frequency-space Action Sequence Tokenization (FAST), enables us to train autoregressive VLAs for highly dexterous and high-frequency tasks where standard discretization methods fail completely. Based on FAST, we release FAST+, a universal robot action tokenizer, trained on 1M real robot action trajectories. It can be used as a black-box tokenizer for a wide range of robot action sequences, with diverse action spaces and control frequencies. Finally, we show that, when combined with the pi0 VLA, our method can scale to training on 10k hours of robot data and match the performance of diffusion VLAs, while reducing training time by up to 5x.

Key facts. Cited by 643 papers in the map (data through 7 Oct 2026). Placed in Efficient VLA inference and adaptation, one of 13 directions in Vision-language-action models. Builds on 37 of its 75 references that are in the map. First posted on arXiv 16 Jan 2025.

Builds on

37 of 75 references.

References · 37

  1. 2017Neural Discrete Representation Learningcited by 4,650
  2. 2023RT-2: Vision-Language-Action Models Transfer Web Knowledge to Robotic Controlcited by 2,993
  3. 2015Neural Machine Translation of Rare Words with Subword Unitscited by 2,898
  4. 2024OpenVLA: An Open-Source Vision-Language-Action Modelcited by 2,777
  5. 2022Do As I Can, Not As I Say: Grounding Language in Robotic Affordancescited by 2,302
  6. 2024π0: A Vision-Language-Action Flow Model for General Robot Controlcited by 2,297
  7. 2023PaLM-E: An Embodied Multimodal Language Modelcited by 2,037
  8. 2022RT-1: Robotics Transformer for Real-World Control at Scalecited by 2,007
  9. 2017Decoupled Weight Decay Regularizationcited by 1,722
  10. 2023Learning Fine-Grained Bimanual Manipulation with Low-Cost Hardwarecited by 1,490
  11. 2023Open X-Embodiment: Robotic Learning Datasets and RT-X Modelscited by 849
  12. 2021SoundStream: An End-to-End Neural Audio Codeccited by 756
25 more
  1. 2024PaliGemma: A versatile 3B VLM for transfercited by 679
  2. 2023Finite Scalar Quantization: VQ-VAE Made Simplecited by 425
  3. 2022BEATs: Audio Pre-Training with Acoustic Tokenizerscited by 378
  4. 2022MAGVIT: Masked Generative Video Transformercited by 341
  5. 2024TinyVLA: Towards Fast, Data-Efficient Vision-Language-Action Models for Robotic Manipulationcited by 331
  6. 20243D-VLA: A 3D Vision-Language-Action Generative World Modelcited by 320
  7. 2024ReKep: Spatio-Temporal Reasoning of Relational Keypoint Constraints for Robotic Manipulationcited by 305
  8. 2024Latent Action Pretraining from Videoscited by 302
  9. 2024TraceVLA: Visual Trace Prompting Enhances Spatial-Temporal Awareness for Generalist Robotic Policiescited by 298
  10. 2024GR-2: A Generative Video-Language-Action Model with Web-Scale Knowledge for Robot Manipulationcited by 290
  11. 2024RT-H: Action Hierarchies Using Languagecited by 204
  12. 2024NaVILA: Legged Robot Vision-Language-Action Model for Navigationcited by 201
  13. 2024Behavior Generation with Latent Actionscited by 159
  14. 2024Open-TeleVision: Teleoperation with Immersive Active Visual Feedbackcited by 137
  15. 2024ALOHA Unleashed: A Simple Recipe for Robot Dexteritycited by 128
  16. 2024DexMimicGen: Automated Data Generation for Bimanual Dexterous Manipulation via Imitation Learningcited by 104
  17. 2022In-Hand Object Rotation via Rapid Motor Adaptationcited by 102
  18. 2024Learning Visuotactile Skills with Two Multifingered Handscited by 71
  19. 2024QueST: Self-Supervised Skill Abstractions for Learning Continuous Controlcited by 51
  20. 2024ElasticTok: Adaptive Tokenization for Image and Videocited by 44
  21. 2025Beyond Sight: Finetuning Generalist Robot Policies with Heterogeneous Sensors via Language Groundingcited by 40
  22. 2024Hand-Object Interaction Pretraining from Videoscited by 38
  23. 2015Multilingual Language Processing From Bytescited by 26
  24. 2024Bridging the Human to Robot Dexterity Gap through Object-Oriented Rewardscited by 25
  25. 2024Efficient Long Video Tokenization via Coordinate-based Patch Reconstructioncited by 6

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2025: 238; 2026: 405.

643 papers.

Sort:

By area:

  1. 2025π0.5: a Vision-Language-Action Model with Open-World Generalizationcited by 1,653
  2. 2025Fine-Tuning Vision-Language-Action Models: Optimizing Speed and Successcited by 844
  3. 2025SmolVLA: A Vision-Language-Action Model for Affordable and Efficient Roboticscited by 507
  4. 2025SpatialVLA: Exploring Spatial Representations for Visual-Language-Action Modelcited by 502
  5. 2025WorldVLA: Towards Autoregressive Action World Modelcited by 351
  6. 2026Causal World Modeling for Robot Controlcited by 289
  7. 2025X-VLA: Soft-Prompted Transformer as Scalable Cross-Embodiment Vision-Language-Action Modelcited by 285
  8. 2025π*0.6: a VLA That Learns From Experiencecited by 280
  9. 2025LIBERO-Plus: In-depth Robustness Analysis of Vision-Language-Action Modelscited by 278
  10. 2025AutoVLA: A Vision-Language-Action Model for End-to-End Autonomous Driving with Adaptive Reasoning and Reinforcement Fine-Tuningcited by 252
  11. 2025Hi Robot: Open-Ended Instruction Following with Hierarchical Vision-Language-Action Modelscited by 216
  12. 2025DreamVLA: A Vision-Language-Action Model Dreamed with Comprehensive World Knowledgecited by 203
38 more
  1. 2025Real-Time Execution of Action Chunking Flow Policiescited by 199
  2. 2025DexVLA: Vision-Language Model with Plug-In Diffusion Expert for General Robot Controlcited by 191
  3. 2025MolmoAct: Action Reasoning Models that can Reason in Spacecited by 189
  4. 2025MemoryVLA: Perceptual-Cognitive Memory in Vision-Language-Action Models for Robotic Manipulationcited by 183
  5. 2025HybridVLA: Collaborative Diffusion and Autoregression in a Unified Vision-Language-Action Modelcited by 176
  6. 2025VLA-RL: Towards Masterful and General Robotic Manipulation with Scalable Reinforcement Learningcited by 166
  7. 2025Ctrl-World: A Controllable Generative World Model for Robot Manipulationcited by 149
  8. 2025Knowledge Insulating Vision-Language-Action Models: Train Fast, Run Fast, Generalize Bettercited by 142
  9. 2025SimpleVLA-RL: Scaling VLA Training via Reinforcement Learningcited by 141
  10. 2025VLA-Adapter: An Effective Paradigm for Tiny-Scale Vision-Language-Action Modelcited by 140
  11. 2025mimic-video: Video-Action Models for Generalizable Robot Control Beyond VLAscited by 133
  12. 2025Interactive Post-Training for Vision-Language-Action Modelscited by 129
  13. 2025Unified Vision-Language-Action Modelcited by 128
  14. 2025DriveVLA-W0: World Models Amplify Data Scaling Law in Autonomous Drivingcited by 127
  15. 2026π0.7: a Steerable Generalist Robotic Foundation Model with Emergent Capabilitiescited by 127
  16. 2025Spatial Forcing: Implicit Spatial Representation Alignment for Vision-language-action Modelcited by 125
  17. 2025NORA: A Small Open-Sourced Generalist Vision Language Action Model for Embodied Taskscited by 119
  18. 2025ChatVLA: Unified Multimodal Understanding and Robot Control with Vision-Language-Action Modelcited by 112
  19. 2025GR-3 Technical Reportcited by 110
  20. 2025OneTwoVLA: A Unified Vision-Language-Action Model with Adaptive Reasoningcited by 105
  21. 2025FLARE: Robot Learning with Implicit World Modelingcited by 105
  22. 2026VLA-JEPA: Enhancing Vision-Language-Action Model with Latent World Modelcited by 104
  23. 2026StarVLA: A Lego-like Codebase for Vision-Language-Action Model Developingcited by 103
  24. 2025PointVLA: Injecting the 3D World into Vision-Language-Action Modelscited by 101
  25. 2025A Survey on Vision-Language-Action Models: An Action Tokenization Perspectivecited by 96
  26. 2025Being-H0: Vision-Language-Action Pretraining from Large-Scale Human Videoscited by 94
  27. 2026ABot-M0: VLA Foundation Model for Robotic Manipulation with Action Manifold Learningcited by 94
  28. 2025Discrete Diffusion VLA: Bringing Discrete Diffusion to Action Decoding in Vision-Language-Action Policiescited by 90
  29. 2025Galaxea Open-World Dataset and G0 Dual-System VLA Modelcited by 90
  30. 2025PD-VLA: Accelerating Vision-Language-Action Model Integrated with Action Chunking via Parallel Decodingcited by 88
  31. 2025ForceVLA: Enhancing VLA Models with a Force-aware MoE for Contact-rich Manipulationcited by 87
  32. 2025F1: A Vision-Language-Action Model Bridging Understanding and Generation to Actionscited by 85
  33. 2025VLA-Cache: Efficient Vision-Language-Action Manipulation via Adaptive Token Cachingcited by 83
  34. 2025InternVLA-M1: A Spatially Guided Vision-Language-Action Framework for Generalist Robot Policycited by 83
  35. 2025Vision-Language-Action Models for Robotics: A Review Towards Real-World Applicationscited by 82
  36. 2025SAFE: Multitask Failure Detection for Vision-Language-Action Modelscited by 74
  37. 2025Vision-Language-Action (VLA) Models: Concepts, Progress, Applications and Challengescited by 72
  38. 2025A Survey on Vision-Language-Action Models for Autonomous Drivingcited by 72

50 most cited of 643 All 643 (JSON)

Similar, not cited

Close in wording, no citation either way

  1. 2026Beyond Reconstruction: What Matters in Action Tokenization for Robot Policies?
  2. 2026ActionCodec: What Makes for Good Action Tokenizers
  3. 2026ActionPiece: Rethinking Action Tokenization for Autoregressive Vision-Language-Action Models
  4. 2025Enhancing Generalization in Vision-Language-Action Models by Preserving Pretrained Representations
  5. 2026Reshaping Action Error Distributions for Reliable Vision-Language-Action Models
  6. 2024CogACT: A Foundational Vision-Language-Action Model for Synergizing Cognition and Action in Robotic Manipulation
  7. 2025XR-1: Towards Versatile Vision-Language-Action Models via Learning Unified Vision-Motion Representations
  8. 2026QVLA: Not All Channels Are Equal in Vision-Language-Action Model's Quantization