mlsanityData to 7 Oct 2026

VLA architectures and generalization

43 papers3 this weekHigh confidence in this grouping: the papers form one clear topicHigh confidence in this grouping: the papers form one clear topic · About

Keywords: vla, action, visual, vision-language-action, language

Key facts. VLA architectures and generalization: 43 papers on arXiv since 2024 (data through 7 Oct 2026). 4 of its 43 mapped papers (9%) cite “Don't Blind Your VLA: Aligning Visual Representations for OOD Generalization” (2025, arXiv 2510.25616). 3 of its 43 mapped papers (7%) cite “A Survey on Vision-Language-Action Models: An Action Tokenization Perspective” (2025, arXiv 2507.01925). 13 new papers 17 Sep–7 Oct.

What’s new

17 Sep – 7 Oct 2026

13 new papers, 17 Sep – 7 Oct; 3 this week.

  1. This week · 3
  2. Do Vision-Language-Action Models Understand Instructions? A Mechanistic Interpretability Study on Language Grounding
  3. DIVA: Dual-Space Intent-Aware Visual Attenuation for Vision-Language-Action Policies also fits: Efficient VLA inference and adaptation
  4. PAIR: Bridging Perception and Action in Vision-Language-Action Models also fits: Generalist VLA models and benchmarks
  5. Earlier · 10
  6. PermVLA: Factorization Order as a Regularizer for VLA Learning
  7. Continuous Conditioning of VLAs with Augmenting EMG and Visual Task Descriptors
  8. Disentangling Spurious Correlations in Vision-Language-Action Models via Predicting Domain-Invariant Latent Lookahead
  9. LexiconVLA: Learning Reusable Atomic Action Codebooks for Unseen Tasks also fits: Reasoning and chain-of-thought in VLA
  10. The Layer Mystery of VLA: An Information-Theoretical Analysis of VLA Latent Interface
  11. Does Adversarial Training Improve Generalization in Multi-View VLAs? Revealing and Mitigating View Collapse also fits: Generalist VLA models and benchmarks
  12. Train Together or Merge Later? Unifying VLA Experts via a Shared Action Interface
All 13
  1. The Linear Representation Hypothesis for Vision-Language-Action Models
  2. CereVLA: Cerebellum-Inspired Consequence-Aware Residual Governance for Efficient Vision-Language-Action Execution also fits: Efficient VLA inference and adaptation
  3. VLAQuantBench: Closed-Loop Evaluation of Post-Training Quantization for Vision-Language-Action Models

Foundations

Most cited by papers here

  1. Don't Blind Your VLA: Aligning Visual Representations for OOD Generalization2025cited by 4 here
  2. A Survey on Vision-Language-Action Models: An Action Tokenization Perspective2025cited by 3 here
  3. Emergent World Representations in OpenVLA2025cited by 3 here
  4. VLM4VLA: Revisiting Vision-Language-Models in Vision-Language-Action Models2026cited by 3 here
  5. Seeing to Act, Prompting to Specify: A Bayesian Factorization of Vision Language Action Policy2025cited by 2 here

Lineage

The papers this direction builds on, oldest first.

Each paper on the line cites the one before.

  1. 2023RT-2: Vision-Language-Action Models Transfer Web Knowledge to Robotic Controlcited by 32 here
  2. 2024Octo: An Open-Source Generalist Robot Policycited by 21 here
  3. 2024OpenVLA: An Open-Source Vision-Language-Action Modelcited by 35 here
  4. 2024TinyVLA: Towards Fast, Data-Efficient Vision-Language-Action Models for Robotic Manipulationcited by 8 here
  5. 2024π0: A Vision-Language-Action Flow Model for General Robot Controlcited by 29 here
  6. 2025FAST: Efficient Action Tokenization for Vision-Language-Action Modelscited by 20 here
  7. 2025OTTER: A Vision-Language-Action Model with Text-Aware Visual Feature Extractioncited by 5 here
  8. 2025π0.5: a Vision-Language-Action Model with Open-World Generalizationcited by 26 here
  9. 2025Knowledge Insulating Vision-Language-Action Models: Train Fast, Run Fast, Generalize Bettercited by 7 here
  10. 2025X-VLA: Soft-Prompted Transformer as Scalable Cross-Embodiment Vision-Language-Action Modelcited by 7 here
14 more it builds on, not in this chain
  1. 20243D-VLA: A 3D Vision-Language-Action Generative World Modelcited by 7 here
  2. 2024Robotic Control via Embodied Chain-of-Thought Reasoningcited by 6 here
  3. 2024CogACT: A Foundational Vision-Language-Action Model for Synergizing Cognition and Action in Robotic Manipulationcited by 8 here
  4. 2024TraceVLA: Visual Trace Prompting Enhances Spatial-Temporal Awareness for Generalist Robotic Policiescited by 5 here
  5. 2024What Matters in Building Vision-Language-Action Models for Generalist Robotscited by 5 here
  6. 2025SpatialVLA: Exploring Spatial Representations for Visual-Language-Action Modelcited by 8 here
  7. 2025Fine-Tuning Vision-Language-Action Models: Optimizing Speed and Successcited by 23 here
  8. 2025CoT-VLA: Visual Chain-of-Thought Reasoning for Vision-Language-Action Modelscited by 11 here
  9. 2025UniVLA: Learning to Act Anywhere with Task-centric Latent Actionscited by 6 here
  10. 2025SmolVLA: A Vision-Language-Action Model for Affordable and Efficient Roboticscited by 13 here
  11. 2025DreamVLA: A Vision-Language-Action Model Dreamed with Comprehensive World Knowledgecited by 5 here
  12. 2025VLA-Adapter: An Effective Paradigm for Tiny-Scale Vision-Language-Action Modelcited by 6 here
  13. 2025LIBERO-PRO: Towards Robust and Fair Evaluation of Vision-Language-Action Models Beyond Memorizationcited by 8 here
  14. 2025LIBERO-Plus: In-depth Robustness Analysis of Vision-Language-Action Modelscited by 10 here

The 24 most cited of 252.

Browse

30 of 43, ranked 6–35

  1. 6VLANeXt: Recipes for Building Strong VLA Models20261
  2. 7Observing and Controlling Features in Vision-Language-Action Models20261
  3. 8Not All Features Are Created Equal: A Mechanistic Study of Vision-Language-Action Models20261
  4. 9VLA-Trace: Diagnosing Vision-Language-Action Models through Representation and Behavior Tracing20261
  5. 10NORA-1.5: A Vision-Language-Action Model Trained using World Model- and Action-based Preference Rewards20251
  6. 11MAPS: Preserving Vision-Language Representations via Module-Wise Proximity Scheduling for Better Vision-Language-Action Generalization20251
  7. 12OmniJARVIS: Unified Vision-Language-Action Tokenization Enables Open-World Instruction Following Agents20241
  8. 13MAIN-VLA: Modeling Abstraction of Intention and eNvironment for Vision-Language-Action Models20260
  9. 14From Representational Complementarity to Dual Systems: Synergizing VLM and Vision-Only Backbones for End-to-End Driving20260
  10. 15LangGap: Diagnosing and Closing the Language Gap in Vision-Language-Action Models20260
All 30
  1. 16vla-eval: A Unified Evaluation Harness for Vision-Language-Action Models20260
  2. 17FocusVLA: Focused Visual Utilization for Vision-Language-Action Models20260
  3. 18VLA-InfoEntropy: A Training-Free Vision-Attention Information Entropy Approach for Vision-Language-Action Models Inference Acceleration and Success20260
  4. 192D or 3D: Who Governs Salience in VLA Models? -- Tri-Stage Token Pruning Framework with Modality Salience Awareness20260
  5. 20VLA Foundry: A Unified Framework for Training Vision-Language-Action Models20260
  6. 21Breaking Lock-In: Preserving Steerability under Low-Data VLA Post-Training20260
  7. 22Embodied Interpretability: Linking Causal Understanding to Generalization in Vision-Language-Action Models20260
  8. 23One Token Per Frame: Reconsidering Visual Bandwidth in World Models for VLA Policy20260
  9. 24CapVector: Learning Transferable Capability Vectors in Parametric Space for Vision-Language-Action Models20260
  10. 25StableVLA: Towards Robust Vision-Language-Action Models without Extra Data20260
  11. 26CrossVLA: Cross-Paradigm Post-Training and Inference Optimization for Vision-Language-Action Models20260
  12. 27Pre-VLA: Preemptive Runtime Verification for Reliable Vision-Language-Action and World-Model Rollouts20260
  13. 28Rethinking VLM Representation for VLA Initialization20260
  14. 29See Less, Specify More: Visual Evidence Budgets for Generalizable VLAs20260
  15. 30FiberTune: Preserving Action-Fiber Visual Residuals in Vision-Language-Action Fine-Tuning20260
  16. 31APT: Action Expert Pretraining Improves Instruction Generalization of Vision-Language-Action Policies20260
  17. 32Revisiting Parameter Redundancy in Vision-Language-Action Models: Insights from VLM-to-VLA Adaptation20260
  18. 33Look Before You Leap: Distilling Tree Search into Action Evaluation for Frozen VLA Models20260
  19. 34Action QFormer: Structured Representation Shaping under Action Supervision in Vision-Language-Action Models20260
  20. 35FabriMAE I Trust Myself? Self-Evaluating VLA Action Generation with Markov Attention Entropy20260