mlsanityData to 14 Oct 2026

Spatial Forcing: Implicit Spatial Representation Alignment for Vision-language-action Model

arXiv 2510.12276 ↗First posted 14 Oct 2025cs.ROcited by 134

On the map: Vision-language-action models → 3D and spatial-aware VLA

Abstract (arXiv)

Vision-language-action (VLA) models have recently shown strong potential in enabling robots to follow language instructions and execute precise actions. However, most VLAs are built upon vision-language models pretrained solely on 2D data, which lack accurate spatial awareness and hinder their ability to operate in the 3D physical world. Existing solutions attempt to incorporate explicit 3D sensor inputs such as depth maps or point clouds, but these approaches face challenges due to sensor noise, hardware heterogeneity, and incomplete depth coverage in existing datasets. Alternative methods that estimate 3D cues from 2D images also suffer from the limited performance of depth estimators. We propose Spatial Forcing (SF), a simple yet effective alignment strategy that implicitly forces VLA models to develop spatial comprehension capabilities without relying on explicit 3D inputs or depth estimators. SF aligns intermediate visual embeddings of VLAs with geometric representations produced by pretrained 3D foundation models. By enforcing alignment at intermediate layers, SF guides VLAs to encode richer spatial representations that enhance action precision. Extensive experiments in simulation and real-world environments demonstrate that SF achieves state-of-the-art results, surpassing both 2D- and 3D-based VLAs. SF further accelerates training by up to 3.8x and improves data efficiency across diverse robotic tasks. Project page is at https://spatial-forcing.github.io/

Key facts. Cited by 134 papers in the map (data through 14 Oct 2026). Placed in 3D and spatial-aware VLA, one of 13 directions in Vision-language-action models. Builds on 79 of its 84 references that are in the map. First posted on arXiv 14 Oct 2025.

Builds on

79 of 84 references.

Closest ancestors · 9

Same direction, newest first

  1. 2025GeoVLA: Empowering 3D Representations in Vision-Language-Action Models12 Aug
  2. 2025Evo-0: Vision-Language-Action Model with Implicit Spatial Understanding1 Jul
  3. 20254D-VLA: Spatiotemporal Vision-Language-Action Pretraining with Cross-Scene Calibration27 Jun
  4. 2025BridgeVLA: Input-Output Alignment for Efficient 3D Manipulation Learning with Vision-Language Models9 Jun
  5. 20253D CAVLA: Leveraging Depth and 3D Context to Generalize Vision Language Action Models for Unseen Tasks9 May
  6. 2025PointVLA: Injecting the 3D World into Vision-Language-Action Models10 Mar
  7. 2025SpatialVLA: Exploring Spatial Representations for Visual-Language-Action Model27 Jan
  8. 2024TraceVLA: Visual Trace Prompting Enhances Spatial-Temporal Awareness for Generalist Robotic Policies13 Dec
  9. 20243D-VLA: A 3D Vision-Language-Action Generative World Model14 Mar

Other references · 70

  1. 2020Denoising Diffusion Probabilistic Modelscited by 21,076
  2. 2023Llama 2: Open Foundation and Fine-Tuned Chat Modelscited by 13,028
  3. 2023DINOv2: Learning Robust Visual Features without Supervisioncited by 7,974
  4. 2023Visual Instruction Tuningcited by 7,869
  5. 2019Momentum Contrast for Unsupervised Visual Representation Learningcited by 7,513
  6. 2022Scalable Diffusion Models with Transformerscited by 6,054
  7. 2022Flow Matching for Generative Modelingcited by 5,702
  8. 2023Improved Baselines with Visual Instruction Tuningcited by 4,418
  9. 2023Sigmoid Loss for Language Image Pre-Trainingcited by 3,109
  10. 2023RT-2: Vision-Language-Action Models Transfer Web Knowledge to Robotic Controlcited by 3,080
  11. 2024OpenVLA: An Open-Source Vision-Language-Action Modelcited by 2,887
  12. 2023Diffusion Policy: Visuomotor Policy Learning via Action Diffusioncited by 2,838
38 more
  1. 2024π0: A Vision-Language-Action Flow Model for General Robot Controlcited by 2,407
  2. 2017FiLM: Visual Reasoning with a General Conditioning Layercited by 2,333
  3. 2022RT-1: Robotics Transformer for Real-World Control at Scalecited by 2,061
  4. 2021Vision Transformers for Dense Predictioncited by 1,803
  5. 2025VGGT: Visual Geometry Grounded Transformercited by 1,539
  6. 2024Octo: An Open-Source Generalist Robot Policycited by 1,484
  7. 2023LIBERO: Benchmarking Knowledge Transfer for Lifelong Robot Learningcited by 1,410
  8. 2025GR00T N1: An Open Foundation Model for Generalist Humanoid Robotscited by 1,225
  9. 2023Self-Supervised Learning from Images with a Joint-Embedding Predictive Architecturecited by 1,039
  10. 2025Fine-Tuning Vision-Language-Action Models: Optimizing Speed and Successcited by 883
  11. 2023VoxPoser: Composable 3D Value Maps for Robotic Manipulation with Language Modelscited by 750
  12. 2021What Matters in Learning from Offline Human Demonstrations for Robot Manipulationcited by 725
  13. 2024Representation Alignment for Generation: Training Diffusion Transformers Is Easier Than You Thinkcited by 703
  14. 2024PaliGemma: A versatile 3B VLM for transfercited by 696
  15. 2024RDT-1B: a Diffusion Foundation Model for Bimanual Manipulationcited by 691
  16. 2025FAST: Efficient Action Tokenization for Vision-Language-Action Modelscited by 665
  17. 2023BridgeData V2: A Dataset for Robot Learning at Scalecited by 657
  18. 2021CALVIN: A Benchmark for Language-Conditioned Policy Learning for Long-Horizon Robot Manipulation Taskscited by 584
  19. 2022BC-Z: Zero-Shot Task Generalization with Robotic Imitation Learningcited by 523
  20. 2023CoTracker: It is Better to Track Togethercited by 522
  21. 2024Evaluating Real-World Robot Manipulation Policies in Simulationcited by 495
  22. 2025CoT-VLA: Visual Chain-of-Thought Reasoning for Vision-Language-Action Modelscited by 451
  23. 2024CogACT: A Foundational Vision-Language-Action Model for Synergizing Cognition and Action in Robotic Manipulationcited by 422
  24. 2025UniVLA: Learning to Act Anywhere with Task-centric Latent Actionscited by 410
  25. 2019Relay Policy Learning: Solving Long-Horizon Tasks via Imitation and Reinforcement Learningcited by 384
  26. 2024Prismatic VLMs: Investigating the Design Space of Visually-Conditioned Language Modelscited by 339
  27. 2024TinyVLA: Towards Fast, Data-Efficient Vision-Language-Action Models for Robotic Manipulationcited by 339
  28. 2023ManiSkill2: A Unified Benchmark for Generalizable Manipulation Skillscited by 327
  29. 2022Behavior Transformers: Cloning k modes with one stonecited by 304
  30. 2024VILA-U: a Unified Foundation Model Integrating Visual Understanding and Generationcited by 295
  31. 2024PaliGemma 2: A Family of Versatile VLMs for Transfercited by 207
  32. 2023Open-World Object Manipulation using Pre-trained Vision-Language Modelscited by 156
  33. 2025VLA-Adapter: An Effective Paradigm for Tiny-Scale Vision-Language-Action Modelcited by 145
  34. 2024ALOHA Unleashed: A Simple Recipe for Robot Dexteritycited by 134
  35. 2023PaLI-3 Vision Language Models: Smaller, Faster, Strongercited by 126
  36. 2025GraspVLA: a Grasping Foundation Model Pre-trained on Billion-scale Synthetic Action Datacited by 123
  37. 2025RoboTwin: Dual-Arm Robot Benchmark with Generative Digital Twinscited by 121
  38. 2019Self-Supervised Correspondence in Visuomotor Policy Learningcited by 107

50 most cited of 70 All 70 (JSON)

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2025: 4; 2026: 130.

134 papers.

Sort:

By area:

  1. 2026A Pragmatic VLA Foundation Modelcited by 103
  2. 2026Being-H0.7: A Latent World-Action Model from Egocentric Videoscited by 75
  3. 2026Unified 4D World Action Modeling from Video Priors with Asynchronous Denoisingcited by 71
  4. 2026RoboDojo: A Unified Sim-and-Real Benchmark for Comprehensive Evaluation of Generalist Robot Manipulation Policiescited by 40
  5. 2026ImageWAM: Do World Action Models Really Need Video Generation, or Just Image Editing?cited by 38
  6. 2025HiF-VLA: Hindsight, Insight and Foresight through Motion Representation for Vision-Language-Action Modelscited by 32
  7. 2026When Vision Overrides Language: Evaluating and Mitigating Counterfactual Failures in VLAscited by 30
  8. 2026Xiaomi-Robotics-1: Scaling Vision-Language-Action Models with over 100K Hours of Real-World Trajectoriescited by 29
  9. 2026LaST-VLA: Thinking in Latent Spatio-Temporal Space for Vision-Language-Action in Autonomous Drivingcited by 28
  10. 2026VGGT-Omegacited by 26
  11. 2026AtomVLA: Scalable Post-Training for Robotic Manipulation via Predictive Latent World Modelscited by 23
  12. 2025Mixture of Horizons in Action Chunkingcited by 21
38 more
  1. 2026VGGDrive: Empowering Vision-Language Models with Cross-View Geometric Grounding for Autonomous Drivingcited by 20
  2. 2026FRAPPE: Infusing World Modeling into Generalist Policies via Multiple Future Representation Alignmentcited by 19
  3. 2026Thinking with Geometry: Active Geometry Integration for Spatial Reasoningcited by 18
  4. 2026S-VAM: Shortcut Video-Action Model by Self-Distilling Geometric and Semantic Foresightcited by 17
  5. 2026FutureVLA: Joint Visuomotor Prediction for Vision-Language-Action Modelcited by 16
  6. 2026Evo-Depth: A Lightweight Depth-Enhanced Vision-Language-Action Modelcited by 16
  7. 2026WAM4D: Fast 4D World Action Model via Spatial Register Tokenscited by 15
  8. 2026TwinBrainVLA: Unleashing the Potential of Generalist VLMs for Embodied Tasks via Asymmetric Mixture-of-Transformerscited by 14
  9. 2026Any3D-VLA: Enhancing VLA Robustness via Diverse Point Cloudscited by 13
  10. 2026VGGT-World: Transforming VGGT into an Autoregressive Geometry World Modelcited by 13
  11. 20264D-WAM: Infusing Spatiotemporal Awareness into World Action Models through Trajectory Fieldscited by 13
  12. 2026AffordanceVLA: A Vision-Language-Action Model Empowering Action Generation through Affordance-Aware Understandingcited by 12
  13. 2025GLaD: Geometric Latent Distillation for Vision-Language-Action Modelscited by 11
  14. 2026ROCKET: Residual-Oriented Multi-Layer Alignment for Spatially-Aware Vision-Language-Action Modelscited by 11
  15. 2026Geometric Action Model for Robot Policy Learningcited by 11
  16. 2025Point What You Mean: Visually Grounded Instruction Policycited by 10
  17. 2026GuidedVLA: Specifying Task-Relevant Factors via Plug-and-Play Action Attention Specializationcited by 10
  18. 2026Rethinking the Practicality of Vision-language-action Model: A Comprehensive Benchmark and An Improved Baselinecited by 9
  19. 2026Pri4R: Learning World Dynamics for Vision-Language-Action Models with Privileged 4D Representationcited by 8
  20. 2026SpatialVAM:Spatial-Aware Multi-View Video Diffusion as a Data-Efficient Robot Policycited by 8
  21. 2026DreamWAM: Beyond RGB Future Prediction for World Action Modelscited by 8
  22. 2026PoseVLA: Universal Pose Pretraining for Generalizable Vision-Language-Action Policiescited by 7
  23. 2026IntentVLA: Short-Horizon Intent Modeling for Aliased Robot Manipulationcited by 7
  24. 2026Revisiting Embodied Chain-of-Thought for Generalizable Robot Manipulationcited by 7
  25. 2026OmniGuide: Universal Guidance Fields for Enhancing Generalist Robot Policiescited by 6
  26. 20263D-Mix for VLA: A Plug-and-Play Module for Integrating VGGT-based 3D Information into Vision-Language-Action Modelscited by 6
  27. 2026FocusVLA: Focused Visual Utilization for Vision-Language-Action Modelscited by 6
  28. 2026PokeVLA: Empowering Pocket-Sized Vision-Language-Action Model with Comprehensive World Knowledge Guidancecited by 6
  29. 2026Breaking Lock-In: Preserving Steerability under Low-Data VLA Post-Trainingcited by 6
  30. 2026GEM: Generative Supervision Helps Embodied Intelligencecited by 6
  31. 2026Beyond Data Scaling: Representation-Centric Continued Pre-training for Vision-Language-Action Modelscited by 6
  32. 2026World-Action Models for Robot Learning and Control: A Surveycited by 6
  33. 2026UAOR: Uncertainty-aware Observation Reinjection for Vision-Language-Action Modelscited by 5
  34. 2026AffordVLA: Injecting Affordance Representations into Vision-Language-Action Models via Implicit Feature Alignmentcited by 5
  35. 2026UniTacVLA: Unified Tactile Understanding and Prediction in Vision Language Action Modelscited by 5
  36. 2026Track4Action: Distilling World-Centric 3D Tracker into Vision-Language-Action Policiescited by 5
  37. 2026From Pixels to Tokens: A Systematic Study of Latent Action Supervision for Vision-Language-Action Modelscited by 4
  38. 2026Spatial Memory for Out-of-Vision Manipulation in Vision-Language-Actioncited by 4

50 most cited of 134 All 134 (JSON)

Similar, not cited

Close in wording, no citation either way

  1. 2025Spatial-Aware VLA Pretraining through Visual-Physical Alignment from Human Videos
  2. 2025From Spatial to Actions: Grounding Vision-Language-Action Model in Spatial Foundation Priors
  3. 2026ST4VLA: Spatially Guided Training for Vision-Language-Action Models
  4. 2026Direct Action-Head Injection of A Grounded 3D Point Unlocks Spatial and Task Generalization
  5. 2025DepthVLA: Enhancing Vision-Language-Action Models with Depth-Aware Spatial Reasoning
  6. 2025ROSA: Harnessing Robot States for Vision-Language and Action Alignment
  7. 2025StereoVLA: Enhancing Vision-Language-Action Models with Stereo Vision
  8. 2026GeomVLA: Unifying Scene, Motion, and Action in 3D