mlsanityData to 7 Oct 2026

BridgeVLA: Input-Output Alignment for Efficient 3D Manipulation Learning with Vision-Language Models

arXiv 2506.07961 ↗First posted 9 Jun 2025cs.RO cs.AIcited by 77

On the map: Multimodal LLMs / vision-language models → Embodied and 3D vision-language models

also in 1 more
Abstract (arXiv)

Recently, leveraging pre-trained vision-language models (VLMs) for building vision-language-action (VLA) models has emerged as a promising approach to effective robot manipulation learning. However, only few methods incorporate 3D signals into VLMs for action prediction, and they do not fully leverage the spatial structure inherent in 3D data, leading to low sample efficiency. In this paper, we introduce BridgeVLA, a novel 3D VLA model that (1) projects 3D inputs to multiple 2D images, ensuring input alignment with the VLM backbone, and (2) utilizes 2D heatmaps for action prediction, unifying the input and output spaces within a consistent 2D image space. In addition, we propose a scalable pre-training method that equips the VLM backbone with the capability to predict 2D heatmaps before downstream policy learning. Extensive experiments show the proposed method is able to learn 3D manipulation efficiently and effectively. BridgeVLA outperforms state-of-the-art baseline methods across three simulation benchmarks. In RLBench, it improves the average success rate from 81.4% to 88.2%. In COLOSSEUM, it demonstrates significantly better performance in challenging generalization settings, boosting the average success rate from 56.7% to 64.0%. In GemBench, it surpasses all the comparing baseline methods in terms of average success rate. In real-robot experiments, BridgeVLA outperforms a state-of-the-art baseline method by 32% on average. It generalizes robustly in multiple out-of-distribution settings, including visual disturbances and unseen instructions. Remarkably, it is able to achieve a success rate of 96.8% on 10+ tasks with only 3 trajectories per task, highlighting its extraordinary sample efficiency. Project Website:https://bridgevla.github.io/

Key facts. Cited by 77 papers in the map (data through 7 Oct 2026). Placed in Embodied and 3D vision-language models, one of 16 directions in Multimodal LLMs / vision-language models. Builds on 46 of its 55 references that are in the map. First posted on arXiv 9 Jun 2025.

Builds on

46 of 55 references.

Closest ancestors · 4

Same direction, newest first

  1. 2025PointVLA: Injecting the 3D World into Vision-Language-Action Models10 Mar
  2. 2024RoboPoint: A Vision-Language Model for Spatial Affordance Prediction for Robotics15 Jun
  3. 2023Vision-Language Foundation Models as Effective Robot Imitators2 Nov
  4. 2023RT-2: Vision-Language-Action Models Transfer Web Knowledge to Robotic Control28 Jul

Other references · 42

  1. 2021Learning Transferable Visual Models From Natural Language Supervisioncited by 33,456
  2. 2023DINOv2: Learning Robust Visual Features without Supervisioncited by 7,764
  3. 2025Qwen2.5-VL Technical Reportcited by 5,418
  4. 2024Qwen2-VL: Enhancing Vision-Language Model's Perception of the World at Any Resolutioncited by 4,151
  5. 2023Sigmoid Loss for Language Image Pre-Trainingcited by 3,048
  6. 2024OpenVLA: An Open-Source Vision-Language-Action Modelcited by 2,777
  7. 2023Diffusion Policy: Visuomotor Policy Learning via Action Diffusioncited by 2,730
  8. 2020RAFT: Recurrent All-Pairs Field Transforms for Optical Flowcited by 2,385
  9. 2024π0: A Vision-Language-Action Flow Model for General Robot Controlcited by 2,297
  10. 2022RT-1: Robotics Transformer for Real-World Control at Scalecited by 2,007
  11. 2025π0.5: a Vision-Language-Action Model with Open-World Generalizationcited by 1,653
  12. 2023Learning Fine-Grained Bimanual Manipulation with Low-Cost Hardwarecited by 1,490
30 more
  1. 2024PaliGemma: A versatile 3B VLM for transfercited by 679
  2. 2022R3M: A Universal Visual Representation for Robot Manipulationcited by 678
  3. 2019RLBench: The Robot Learning Benchmark & Learning Environmentcited by 669
  4. 2021Perceiver IO: A General Architecture for Structured Inputs & Outputscited by 616
  5. 2023Point Transformer V3: Simpler, Faster, Strongercited by 579
  6. 2022Perceiver-Actor: A Multi-Task Transformer for Robotic Manipulationcited by 545
  7. 2022BC-Z: Zero-Shot Task Generalization with Robotic Imitation Learningcited by 517
  8. 2025SpatialVLA: Exploring Spatial Representations for Visual-Language-Action Modelcited by 502
  9. 2022PointNeXt: Revisiting PointNet++ with Improved Training and Scaling Strategiescited by 432
  10. 2023Unleashing Large-Scale Video Generative Pre-training for Visual Robot Manipulationcited by 353
  11. 2024Prismatic VLMs: Investigating the Design Space of Visually-Conditioned Language Modelscited by 336
  12. 2023Scaling Open-Vocabulary Object Detectioncited by 335
  13. 20243D-VLA: A 3D Vision-Language-Action Generative World Modelcited by 320
  14. 2024GR-2: A Generative Video-Language-Action Model with Web-Scale Knowledge for Robot Manipulationcited by 290
  15. 20243D Diffuser Actor: Policy Diffusion with 3D Scene Representationscited by 278
  16. 2022Masked Visual Pre-training for Motor Controlcited by 246
  17. 2023RVT: Robotic View Transformer for 3D Object Manipulationcited by 224
  18. 2023Act3D: 3D Feature Field Transformers for Multi-Task Robotic Manipulationcited by 135
  19. 2024RVT-2: Learning Precise Manipulation from Few Demonstrationscited by 135
  20. 2024THE COLOSSEUM: A Benchmark for Evaluating Generalization for Robotic Manipulationcited by 128
  21. 2021Coarse-to-Fine Q-attention: Efficient Learning for Visual Robotic Manipulation via Discretisationcited by 120
  22. 2022Instruction-driven history-aware policies for robotic manipulationscited by 93
  23. 2021Coarse-to-Fine Imitation Learning: Robot Manipulation from a Single Demonstrationcited by 92
  24. 2014Reducing the Barrier to Entry of Complex Robotic Software: a MoveIt! Case Studycited by 86
  25. 2023PolarNet: 3D Point Clouds for Language-Guided Robotic Manipulationcited by 56
  26. 2024Lift3D Foundation Policy: Lifting 2D Large-Scale Pretrained Models for Robust 3D Robotic Manipulationcited by 45
  27. 2023M2T2: Multi-Task Masked Transformer for Object-centric Pick and Placecited by 40
  28. 2024GR-MG: Leveraging Partially Annotated Data via Multi-Modal Goal-Conditioned Policycited by 39
  29. 2024Towards Generalizable Vision-Language Robotic Manipulation: A Benchmark and LLM-guided 3D Policycited by 37
  30. 2025FP3: A 3D Foundation Policy for Robotic Manipulationcited by 31

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2025: 26; 2026: 51.

77 papers.

Sort:

By area:

  1. 2025Spatial Forcing: Implicit Spatial Representation Alignment for Vision-language-action Modelcited by 125
  2. 2025Vision-Language-Action Models for Robotics: A Review Towards Real-World Applicationscited by 82
  3. 2025Large VLM-based Vision-Language-Action Models for Robotic Manipulation: A Surveycited by 66
  4. 2026Unified 4D World Action Modeling from Video Priors with Asynchronous Denoisingcited by 65
  5. 2025Evo-0: Vision-Language-Action Model with Implicit Spatial Understandingcited by 55
  6. 2025Pure Vision Language Action (VLA) Models: A Comprehensive Surveycited by 46
  7. 2025DepthVLA: Enhancing Vision-Language-Action Models with Depth-Aware Spatial Reasoningcited by 44
  8. 2025FailSafe: Reasoning and Recovery from Failures in Vision-Language-Action Modelscited by 33
  9. 2025Multimodal Spatial Reasoning in the Large Model Era: A Survey and Benchmarkscited by 31
  10. 2026Adaptive Action Chunking at Inference-time for Vision-Language-Action Modelscited by 24
  11. 2026Xiaomi-Robotics-1: Scaling Vision-Language-Action Models with over 100K Hours of Real-World Trajectoriescited by 24
  12. 2025SwiftVLA: Unlocking Spatiotemporal Dynamics for Lightweight VLA Models at Minimal Overheadcited by 23
38 more
  1. 2026BridgeV2W: Bridging Video Generation Models to Embodied World Models via Embodiment Maskscited by 19
  2. 2025AsyncVLA: Asynchronous Flow Matching for Vision-Language-Action Modelscited by 16
  3. 2025StereoVLA: Enhancing Vision-Language-Action Models with Stereo Visioncited by 16
  4. 2025GeoPredict: Leveraging Predictive Kinematics and 3D Gaussian Geometry for Precise VLA Manipulationcited by 15
  5. 2026ActiveVLA: Injecting Active Perception into Vision-Language-Action Models for Precise 3D Robotic Manipulationcited by 14
  6. 2026Any3D-VLA: Enhancing VLA Robustness via Diverse Point Cloudscited by 13
  7. 2026MemoryWAM: Efficient World Action Modeling with Persistent Memorycited by 12
  8. 2025Physical Autoregressive Model for Robotic Manipulation without Action Pretrainingcited by 11
  9. 2025EC-Flow: Enabling Versatile Robotic Manipulation from Action-Unlabeled Videos via Embodiment-Centric Flowcited by 10
  10. 2025StaMo: Unsupervised Learning of Generalizable Robot Motion from Compact State Representationcited by 10
  11. 2025Point What You Mean: Visually Grounded Instruction Policycited by 10
  12. 2026FlowWAM: Optical Flow as a Unified Action Representation for World Action Modelscited by 10
  13. 2026SpatialVAM:Spatial-Aware Multi-View Video Diffusion as a Data-Efficient Robot Policycited by 8
  14. 2026ProcVLM: Learning Procedure-Grounded Progress Rewards for Robotic Manipulationcited by 8
  15. 2026AffordanceVLA: A Vision-Language-Action Model Empowering Action Generation through Affordance-Aware Understandingcited by 8
  16. 2025Survey of Vision-Language-Action Models for Embodied Manipulationcited by 7
  17. 2026Pri4R: Learning World Dynamics for Vision-Language-Action Models with Privileged 4D Representationcited by 7
  18. 2026GuidedVLA: Specifying Task-Relevant Factors via Plug-and-Play Action Attention Specializationcited by 7
  19. 2026RotVLA: Rotational Latent Action for Vision-Language-Action Modelcited by 7
  20. 2026RoboRouter: Training-Free Policy Routing for Robotic Manipulationcited by 6
  21. 2026PokeVLA: Empowering Pocket-Sized Vision-Language-Action Model with Comprehensive World Knowledge Guidancecited by 6
  22. 2026ConsisVLA-4D: Advancing Spatiotemporal Consistency in Efficient 3D-Perception and 4D-Reasoning for Robotic Manipulationcited by 6
  23. 2026PointACT: Vision-Language-Action Models with Multi-Scale Point-Action Interactioncited by 6
  24. 2025Learning to See and Act: Task-Aware Virtual View Exploration for Robotic Manipulationcited by 5
  25. 2026UAOR: Uncertainty-aware Observation Reinjection for Vision-Language-Action Modelscited by 5
  26. 2026UniSteer: Unified Noise Steering for Efficient Human-Guided VLA Adaptationcited by 5
  27. 2026DIM-WAM: World-Action Modeling with Diverse Historical Event Memorycited by 5
  28. 2025NoTVLA: Semantics-Preserving Robot Adaptation via Narrative Action Interfacescited by 4
  29. 2025GAE: Unleashing Physical Potential of VLM with Generalizable Action Expertcited by 4
  30. 2025An Anatomy of Vision-Language-Action Models: From Modules to Milestones and Challengescited by 4
  31. 2026Revisiting Embodied Chain-of-Thought for Generalizable Robot Manipulationcited by 4
  32. 2025Robotic Manipulation via Imitation Learning: Taxonomy, Evolution, Benchmark, and Challengescited by 3
  33. 2025VLA-LPAF: Lightweight Perspective-Adaptive Fusion for Vision-Language-Action to Enable More Unconstrained Robotic Manipulationcited by 3
  34. 2026SA-VLA: Spatially-Aware Flow-Matching for Vision-Language-Action Reinforcement Learningcited by 3
  35. 2026Learning Action Manifold with Multi-view Latent Priors for Robotic Manipulationcited by 3
  36. 2026GaussianDream: A Feed-Forward 3D Gaussian World Model for Robotic Manipulationcited by 3
  37. 2026OASIS: Observation-Action Space Alignment via SE(3) Trajectory Prediction for Robotic Manipulationcited by 3
  38. 2026MA-VLA: Multi-Arm Vision-Language-Action Model for Collaboration and Compositional Generalizationcited by 3

50 most cited of 77 All 77 (JSON)

Similar, not cited

Close in wording, no citation either way

  1. 2026Bridge3D: Enabling Vision-Language-Action Models to See and Act in 3D
  2. 2026SimVLA: A Simple VLA Baseline for Robotic Manipulation
  3. 2025cVLA: Towards Efficient Camera-Space VLAs
  4. 2025HAMSTER: Hierarchical Action Models For Open-World Robot Manipulation
  5. 20253D CAVLA: Leveraging Depth and 3D Context to Generalize Vision Language Action Models for Unseen Tasks
  6. 2025GeoVLA: Empowering 3D Representations in Vision-Language-Action Models
  7. 2026AugVLA-3D: Depth-Driven Feature Augmentation for Vision-Language-Action Models
  8. 2026Explicit Kinematic Guidance from Analytic Concepts for Vision-Language-Action Models