mlsanityData to 7 Oct 2026

DepthVLA: Enhancing Vision-Language-Action Models with Depth-Aware Spatial Reasoning

arXiv 2510.13375 ↗First posted 15 Oct 2025cs.CVcited by 44

On the map: LLM reasoning → Spatial and embodied reasoning

also in 1 more
Abstract (arXiv)

Vision-Language-Action (VLA) models have recently shown impressive generalization and language-guided manipulation capabilities. However, their performance degrades on tasks requiring precise spatial reasoning due to limited spatial reasoning inherited from Vision-Language Models (VLMs). Existing VLAs rely on extensive action-data pretraining to ground VLMs in 3D space, which reduces training efficiency and is still insufficient for accurate spatial understanding. In this work, we present DepthVLA, a simple yet effective VLA architecture that explicitly incorporates spatial awareness through a pretrained depth prediction module. DepthVLA adopts a mixture-of-transformers design that unifies a VLM, a depth transformer, and an action expert with fully shared attentions, forming an end-to-end model with enhanced spatial reasoning. Extensive evaluations in both real-world and simulated environments show that DepthVLA outperforms state-of-the-art approaches, achieving 78.5% vs. 65.0% progress in real-world tasks, 94.9% vs. 93.6% in the LIBERO simulator, and 74.8% vs. 58.8% in the Simpler simulator. Our code will be made publicly available.

Key facts. Cited by 44 papers in the map (data through 7 Oct 2026). Placed in Spatial and embodied reasoning, one of 17 directions in LLM reasoning. Builds on 47 of its 49 references that are in the map. First posted on arXiv 15 Oct 2025.

Builds on

47 of 49 references.

Closest ancestors · 3

Same direction, newest first

  1. 2025MolmoAct: Action Reasoning Models that can Reason in Space11 Aug
  2. 2025CoT-VLA: Visual Chain-of-Thought Reasoning for Vision-Language-Action Models27 Mar
  3. 2025PointVLA: Injecting the 3D World into Vision-Language-Action Models10 Mar

Other references · 44

  1. 2023DINOv2: Learning Robust Visual Features without Supervisioncited by 7,764
  2. 2017ScanNet: Richly-annotated 3D Reconstructions of Indoor Scenescited by 3,088
  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. 2023Diffusion Policy: Visuomotor Policy Learning via Action Diffusioncited by 2,730
  6. 2024π0: A Vision-Language-Action Flow Model for General Robot Controlcited by 2,297
  7. 2023DUSt3R: Geometric 3D Vision Made Easycited by 1,526
  8. 2025VGGT: Visual Geometry Grounded Transformercited by 1,475
  9. 2024Octo: An Open-Source Generalist Robot Policycited by 1,427
  10. 2024Depth Anything V2cited by 1,385
  11. 2024Depth Anything: Unleashing the Power of Large-Scale Unlabeled Datacited by 1,346
  12. 2023LIBERO: Benchmarking Knowledge Transfer for Lifelong Robot Learningcited by 1,338
32 more
  1. 2014Depth Map Prediction from a Single Image using a Multi-Scale Deep Networkcited by 1,264
  2. 2024Grounding Image Matching in 3D with MASt3Rcited by 960
  3. 2024DROID: A Large-Scale In-The-Wild Robot Manipulation Datasetcited by 801
  4. 2023ScanNet++: A High-Fidelity Dataset of 3D Indoor Scenescited by 747
  5. 2024Eyes Wide Shut? Exploring the Visual Shortcomings of Multimodal LLMscited by 719
  6. 2024PaliGemma: A versatile 3B VLM for transfercited by 679
  7. 2023BridgeData V2: A Dataset for Robot Learning at Scalecited by 634
  8. 2020Hypersim: A Photorealistic Synthetic Dataset for Holistic Indoor Scene Understandingcited by 593
  9. 2025SpatialVLA: Exploring Spatial Representations for Visual-Language-Action Modelcited by 502
  10. 2024Evaluating Real-World Robot Manipulation Policies in Simulationcited by 484
  11. 2024MonST3R: A Simple Approach for Estimating Geometry in the Presence of Motioncited by 431
  12. 2024CogACT: A Foundational Vision-Language-Action Model for Synergizing Cognition and Action in Robotic Manipulationcited by 411
  13. 2024Prismatic VLMs: Investigating the Design Space of Visually-Conditioned Language Modelscited by 336
  14. 20243D-VLA: A 3D Vision-Language-Action Generative World Modelcited by 320
  15. 20243D Reconstruction with Spatial Memorycited by 269
  16. 2025Unified World Models: Coupling Video and Action Diffusion for Pretraining on Large Robotic Datasetscited by 218
  17. 2025DreamVLA: A Vision-Language-Action Model Dreamed with Comprehensive World Knowledgecited by 203
  18. 2024MASt3R-SLAM: Real-Time Dense SLAM with 3D Reconstruction Priorscited by 198
  19. 2025UniDepthV2: Universal Monocular Metric Depth Estimation Made Simplercited by 170
  20. 2024Predictive Inverse Dynamics Models are Scalable Learners for Robotic Manipulationcited by 159
  21. 2024Mixture-of-Transformers: A Sparse and Scalable Architecture for Multi-Modal Foundation Modelscited by 151
  22. 2024Vision language models are blind: Failing to translate detailed visual features into wordscited by 144
  23. 2024RGBD Objects in the Wild: Scaling Real-World 3D Object Learning from RGB-D Videoscited by 113
  24. 2025GR-3 Technical Reportcited by 110
  25. 2025UP-VLA: A Unified Understanding and Prediction Model for Embodied Agentcited by 99
  26. 2025Galaxea Open-World Dataset and G0 Dual-System VLA Modelcited by 90
  27. 2025BridgeVLA: Input-Output Alignment for Efficient 3D Manipulation Learning with Vision-Language Modelscited by 77
  28. 2023Multimodal Representation Learning by Alternating Unimodal Adaptationcited by 74
  29. 2024Perception Tokens Enhance Visual Reasoning in Multimodal Language Modelscited by 68
  30. 2023On Uni-Modal Feature Learning in Supervised Multi-Modal Learningcited by 53
  31. 2024Lift3D Foundation Policy: Lifting 2D Large-Scale Pretrained Models for Robust 3D Robotic Manipulationcited by 45
  32. 2025LONG3R: Long Sequence Streaming 3D Reconstructioncited by 45

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2025: 1; 2026: 43.

44 papers.

Sort:

By area:

  1. 2026World Action Models: The Next Frontier in Embodied AIcited by 46
  2. 2026ImageWAM: Do World Action Models Really Need Video Generation, or Just Image Editing?cited by 33
  3. 2026ACoT-VLA: Action Chain-of-Thought for Vision-Language-Action Modelscited by 30
  4. 2026Adaptive Action Chunking at Inference-time for Vision-Language-Action Modelscited by 24
  5. 2026JEPA-VLA: Video Predictive Embedding is Needed for VLA Modelscited by 22
  6. 2026Generation Enhances Understanding in Unified Multimodal Models via Multi-Representation Generationcited by 19
  7. 2026UniDriveVLA: Unifying Understanding, Perception, and Action Planning for Autonomous Drivingcited by 17
  8. 2025StereoVLA: Enhancing Vision-Language-Action Models with Stereo Visioncited by 16
  9. 2026S-VAM: Shortcut Video-Action Model by Self-Distilling Geometric and Semantic Foresightcited by 16
  10. 2026Any3D-VLA: Enhancing VLA Robustness via Diverse Point Cloudscited by 13
  11. 2026Evo-Depth: A Lightweight Depth-Enhanced Vision-Language-Action Modelcited by 13
  12. 2026SimVLA: A Simple VLA Baseline for Robotic Manipulationcited by 10
32 more
  1. 2026GuidedVLA: Specifying Task-Relevant Factors via Plug-and-Play Action Attention Specializationcited by 7
  2. 2026PointACT: Vision-Language-Action Models with Multi-Scale Point-Action Interactioncited by 6
  3. 2026GEM: Generative Supervision Helps Embodied Intelligencecited by 6
  4. 2026DreamWAM: Beyond RGB Future Prediction for World Action Modelscited by 6
  5. 2026FocusVLA: Focused Visual Utilization for Vision-Language-Action Modelscited by 5
  6. 2026Spatial Memory for Out-of-Vision Manipulation in Vision-Language-Actioncited by 4
  7. 2026Learning Panorama-Aware VLA for Mobile Manipulation with Whole-Body Teleoperationcited by 4
  8. 2026UniLACT: Depth-Aware RGB Latent Action Learning for Vision-Language-Action Modelscited by 3
  9. 2026Look Before Acting: Enhancing Vision Foundation Representations for Vision-Language-Action Modelscited by 3
  10. 2026Learning Action Manifold with Multi-view Latent Priors for Robotic Manipulationcited by 3
  11. 2026Afford-VLA: Action-Aligned Visual Planning via Internalized Affordancecited by 3
  12. 2026GesVLA: Gesture-Aware Vision-Language-Action Model Embedded Representationscited by 2
  13. 2026MotionVLA: Injecting Geometric Motion into Vision-Language-Action Modelcited by 2
  14. 2026GEAR-VLA: Learning Geometry-Aware Action Representations for Generalizable Robotic Manipulationcited by 2
  15. 2026Z-1: Efficient Reinforcement Learning for Vision-Language-Action Modelscited by 2
  16. 2026DepthCache: Depth-Guided Training-Free Visual Token Merging for Vision-Language-Action Model Inferencecited by 1
  17. 2026PVI: Plug-in Visual Injection for Vision-Language-Action Modelscited by 1
  18. 2026R3D: Revisiting 3D Policy Learningcited by 1
  19. 2026GeoAlign: Beyond Semantics with State-Guided Spatial Alignment in VLA Modelscited by 1
  20. 20263DThinkVLA: Endowing Vision-Language-Action Models with Latent 3D Priors via 3D-Thinking-Guided Co-trainingcited by 1
  21. 2026See like a Robot: Robot-Centric Pointmaps for VLA Modelscited by 1
  22. 2026GST-VLA: Structured Gaussian Spatial Tokens for 3D Depth-Aware Vision-Language-Action Modelscited by 0
  23. 2026AnchorVLA4D: an Anchor-Based Spatial-Temporal Vision-Language-Action Model for Robotic Manipulationcited by 0
  24. 2026SpatialPoint: Spatial-aware Point Prediction for Embodied Localizationcited by 0
  25. 2026Why MLLMs Struggle to Determine Object Orientationscited by 0
  26. 2026VistaVLA: Geometry- and Semantic-Aware 3D Gaussian-Grounded VLA for Robotic Manipulationcited by 0
  27. 2026SpatioLM: Towards General Physical Spatial Intelligence in Vision-Language Modelscited by 0
  28. 2026Toward Unified Robot Learning: Bridging Representation, Vision-Language-Action, and World Modelscited by 0
  29. 2026ActiveScale: Scaling Active Perception for Robots across Model, Data, and Hardwarecited by 0
  30. 2026PanoFuse: Panorama-Enhanced Vision-Language-Action Learning with Decoupled Semantic-Geometric Routingcited by 0
  31. 2026Spatial Grafting: Grounding 3D Features for Flow-Matching Robot Policiescited by 0
  32. 2026Explicit Geometric Chain-of-Thought for Vision-Language-Action in Autonomous Drivingcited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2025Evo-0: Vision-Language-Action Model with Implicit Spatial Understanding
  2. 2025GeoVLA: Empowering 3D Representations in Vision-Language-Action Models
  3. 2025QDepth-VLA: Quantized Depth Prediction as Auxiliary Supervision for Vision-Language-Action Models
  4. 2026GaussVLA: Geometry-Aware Spatial Reasoning for Vision-Language-Action Model
  5. 2025Spatial-Aware VLA Pretraining through Visual-Physical Alignment from Human Videos
  6. 2026AugVLA-3D: Depth-Driven Feature Augmentation for Vision-Language-Action Models
  7. 20253D CAVLA: Leveraging Depth and 3D Context to Generalize Vision Language Action Models for Unseen Tasks
  8. 20263DVLA: Enhancing Vision-Language-Action Models via 3D Spatial and Instance Understanding