mlsanityData to 7 Oct 2026

V-Link: Recovering Lost Visual Representations in Action DiT for Vision-Language-Action Models

arXiv 2608.25308 ↗First posted 26 Aug 2026cs.CVcited by 0

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

Abstract (arXiv)

Vision-language-action (VLA) models provide a scalable path toward generalist robotic manipulation by integrating visual perception, language understanding, and continuous action control. However, we reveal a critical limitation of VLA architectures: the action expert has limited access to the 3D geometric and 2D semantic information available in VLM features. This accessibility gap weakens perceptual grounding and limits performance on fine-grained robotic manipulation. To address this issue, we propose V-Link, which explicitly recovers visual representations during the vision-language (VL) to action (A) feature transfer. Specifically, V-Link learns complementary Spatial and Semantic Query representations within the VLM and injects them into Action DiT through asymmetric pathways. Semantic Queries complement the original VLM image tokens, whereas Spatial Queries provide dedicated geometric conditioning for spatially grounded action generation. Across LIBERO, LIBERO-Plus, and RoboTwin 2.0, our V-Link improves the average success rate over base model GR00T N1.6 by +1.9%, +31.2%, and +18.8%, respectively. On the AGIBOT A3 Ultra, V-Link further achieves gains of +20% and +24% on two real-world humanoid tasks.

Key facts. No citing papers in the map yet (data through 7 Oct 2026). Placed in 3D and spatial-aware VLA, one of 13 directions in Vision-language-action models. Builds on 31 of its 31 references that are in the map. First posted on arXiv 26 Aug 2026.

Builds on

31 of 31 references.

Closest ancestors · 8

Same direction, newest first

  1. 2026GEAR-VLA: Learning Geometry-Aware Action Representations for Generalizable Robotic Manipulation7 Jun
  2. 2026Evo-Depth: A Lightweight Depth-Enhanced Vision-Language-Action Model14 May
  3. 2025Spatial Forcing: Implicit Spatial Representation Alignment for Vision-language-action Model14 Oct
  4. 2025GeoVLA: Empowering 3D Representations in Vision-Language-Action Models12 Aug
  5. 2025Evo-0: Vision-Language-Action Model with Implicit Spatial Understanding1 Jul
  6. 20253D CAVLA: Leveraging Depth and 3D Context to Generalize Vision Language Action Models for Unseen Tasks9 May
  7. 2025PointVLA: Injecting the 3D World into Vision-Language-Action Models10 Mar
  8. 2025SpatialVLA: Exploring Spatial Representations for Visual-Language-Action Model27 Jan

Other references · 23

  1. 2024OpenVLA: An Open-Source Vision-Language-Action Modelcited by 2,777
  2. 2024π0: A Vision-Language-Action Flow Model for General Robot Controlcited by 2,297
  3. 2025π0.5: a Vision-Language-Action Model with Open-World Generalizationcited by 1,653
  4. 2025VGGT: Visual Geometry Grounded Transformercited by 1,475
  5. 2023LIBERO: Benchmarking Knowledge Transfer for Lifelong Robot Learningcited by 1,338
  6. 2025GR00T N1: An Open Foundation Model for Generalist Humanoid Robotscited by 1,167
  7. 2025Fine-Tuning Vision-Language-Action Models: Optimizing Speed and Successcited by 844
  8. 2025FAST: Efficient Action Tokenization for Vision-Language-Action Modelscited by 643
  9. 2025RoboTwin 2.0: A Scalable Data Generator and Benchmark with Strong Domain Randomization for Robust Bimanual Robotic Manipulationcited by 477
  10. 2025CoT-VLA: Visual Chain-of-Thought Reasoning for Vision-Language-Action Modelscited by 440
  11. 2025LIBERO-Plus: In-depth Robustness Analysis of Vision-Language-Action Modelscited by 278
  12. 2026Cosmos Policy: Fine-Tuning Video Models for Visuomotor Control and Planningcited by 274
11 more
  1. 2025VLA-Adapter: An Effective Paradigm for Tiny-Scale Vision-Language-Action Modelcited by 140
  2. 2025Unified Vision-Language-Action Modelcited by 128
  3. 2026A Pragmatic VLA Foundation Modelcited by 97
  4. 2026ABot-M0: VLA Foundation Model for Robotic Manipulation with Action Manifold Learningcited by 94
  5. 2025VLA-Cache: Efficient Vision-Language-Action Manipulation via Adaptive Token Cachingcited by 83
  6. 2025Dita: Scaling Diffusion Transformer for Generalist Vision-Language-Action Policycited by 73
  7. 2025SP-VLA: A Joint Model Scheduling and Token Pruning Approach for VLA Model Accelerationcited by 56
  8. 2026Xiaomi-Robotics-0: An Open-Sourced Vision-Language-Action Model with Real-Time Executioncited by 42
  9. 2026Masked Depth Modeling for Spatial Perceptioncited by 24
  10. 2026EventVLA: Event-Driven Visual Evidence Memory for Long-Horizon Vision-Language-Action Policiescited by 18
  11. 2026UniFS: Unified Fast-to-Slow Hierarchical Architecture for Vision-Language-Action Modelscited by 1

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2026Look Before Acting: Enhancing Vision Foundation Representations for Vision-Language-Action Models
  2. 2026M2-VLA: Boosting Vision-Language Models for Generalizable Manipulation via Layer Mixture and Meta-Skills
  3. 2026Beyond Data Scaling: Representation-Centric Continued Pre-training for Vision-Language-Action Models
  4. 2025Enhancing Generalization in Vision-Language-Action Models by Preserving Pretrained Representations
  5. 2026LAST: Bridging Vision-Language and Action Manifolds via Gromov-Wasserstein Alignment
  6. 2026G3VLA: Geometric inductive bias for Vision-Language-Action Models
  7. 2024CogACT: A Foundational Vision-Language-Action Model for Synergizing Cognition and Action in Robotic Manipulation
  8. 2026LoopVLA: Learning Sufficiency in Recurrent Refinement for Vision-Language-Action Models