mlsanityData to 7 Oct 2026

Are Vision-Language-Action Models Robust to One-Step Observation Perturbations?

arXiv 2609.32550 ↗First posted 26 Sep 2026cs.AI cs.LG cs.ROcited by 0

On the map: Vision-language-action models → Safety and adversarial attacks

Abstract (arXiv)

Understanding the safety risks of vision-language-action (VLA) models is essential for their deployment in the physical world. Existing safety research has mainly considered persistent perturbations that are applied continuously to observations throughout an episode. However, momentary observation corruption, in which observations are severely perturbed only briefly within an episode, remains an underexplored safety threat. To address this gap, this work investigates robustness to one-step perturbations applied at a single time step per episode. Our experiments reveal that these perturbations substantially degrade VLA performance and that their impact depends on the action chunk execution length. Based on them, we propose CARE, which dynamically selects the execution length based on consistency with the previously predicted action chunk. CARE improves robustness with low computational overhead while preserving clean performance by selecting shorter execution lengths only under perturbations.

Key facts. No citing papers in the map yet (data through 7 Oct 2026). Placed in Safety and adversarial attacks, one of 13 directions in Vision-language-action models. Builds on 37 of its 43 references that are in the map. First posted on arXiv 26 Sep 2026.

Builds on

37 of 43 references.

Closest ancestors · 7

Same direction, newest first

  1. 2026Vision-Language-Action Safety: Threats, Challenges, Evaluations, and Mechanisms26 Apr
  2. 2026STRONG-VLA: Decoupled Robustness Learning for Vision-Language-Action Models under Multimodal Perturbations11 Apr
  3. 2025Phantom Menace: Exploring and Enhancing the Robustness of VLA Models Against Physical Sensor Attacks13 Nov
  4. 2025Model-agnostic Adversarial Attack and Defense for Vision-Language-Action Models15 Oct
  5. 2025RobustVLA: On Robustness of Vision-Language-Action Model against Multi-Modal Perturbations26 Sep
  6. 2025Adversarial Attacks on Robotic Vision Language Action Models3 Jun
  7. 2024Exploring the Adversarial Vulnerabilities of Vision-Language-Action Models in Robotics18 Nov

Other references · 30

  1. 2023RT-2: Vision-Language-Action Models Transfer Web Knowledge to Robotic Controlcited by 2,993
  2. 2024OpenVLA: An Open-Source Vision-Language-Action Modelcited by 2,777
  3. 2023Diffusion Policy: Visuomotor Policy Learning via Action Diffusioncited by 2,730
  4. 2024π0: A Vision-Language-Action Flow Model for General Robot Controlcited by 2,297
  5. 2025π0.5: a Vision-Language-Action Model with Open-World Generalizationcited by 1,653
  6. 2023Learning Fine-Grained Bimanual Manipulation with Low-Cost Hardwarecited by 1,490
  7. 2024Octo: An Open-Source Generalist Robot Policycited by 1,427
  8. 2023LIBERO: Benchmarking Knowledge Transfer for Lifelong Robot Learningcited by 1,338
  9. 2019Meta-World: A Benchmark and Evaluation for Multi-Task and Meta Reinforcement Learningcited by 1,168
  10. 2024RDT-1B: a Diffusion Foundation Model for Bimanual Manipulationcited by 670
  11. 2025FAST: Efficient Action Tokenization for Vision-Language-Action Modelscited by 643
  12. 2025SmolVLA: A Vision-Language-Action Model for Affordable and Efficient Roboticscited by 507
18 more
  1. 2025X-VLA: Soft-Prompted Transformer as Scalable Cross-Embodiment Vision-Language-Action Modelcited by 285
  2. 2025Real-Time Execution of Action Chunking Flow Policiescited by 199
  3. 2025DexVLA: Vision-Language Model with Plug-In Diffusion Expert for General Robot Controlcited by 191
  4. 2024Unpacking Failure Modes of Generative Policies: Runtime Monitoring of Consistency and Progresscited by 63
  5. 2024Bidirectional Decoding: Improving Action Chunking via Guided Test-Time Samplingcited by 46
  6. 2024VLATest: Testing and Evaluating Vision-Language-Action Models for Robotic Manipulationcited by 43
  7. 2024Run-time Observation Interventions Make Vision-Language-Action Models More Visually Robustcited by 33
  8. 2025GEVRM: Goal-Expressive Video Generation Model For Robust Visual Manipulationcited by 26
  9. 2025CronusVLA: Towards Efficient and Robust Manipulation via Multi-Frame Vision-Language-Action Modelingcited by 24
  10. 2026Adaptive Action Chunking at Inference-time for Vision-Language-Action Modelscited by 24
  11. 2025Mixture of Horizons in Action Chunkingcited by 20
  12. 2026When to Trust Imagination: Adaptive Action Execution for World Action Modelscited by 18
  13. 2024HumanVLA: Towards Vision-Language Directed Object Rearrangement by Physical Humanoidcited by 17
  14. 2025RobustVLA: Robustness-Aware Reinforcement Post-Training for Vision-Language-Action Modelscited by 16
  15. 2026VLA-Corrector: Lightweight Detect-and-Correct Inference for Adaptive Action Horizoncited by 14
  16. 2024Benchmarking the Robustness of Temporal Action Detection Models Against Temporal Corruptionscited by 12
  17. 2025Improving Generative Behavior Cloning via Self-Guidance and Adaptive Chunkingcited by 7
  18. 2026VLA Knows Its Limits: Adaptive Execution Horizons for Robot Policiescited by 7

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2025LIBERO-Plus: In-depth Robustness Analysis of Vision-Language-Action Models
  2. 2026HazardArena: Evaluating Semantic Safety in Vision-Language-Action Models
  3. 2026Concept-Based Dictionary Learning for Inference-Time Safety in Vision Language Action Models
  4. 2026A Safe Action Is Not Enough: Feasible-Future Decoding for Vision-Language-Action Policies
  5. 2026CARE: Certifying Acceleration for Vision-Language-Action Inference
  6. 2025A Survey of Safety on Large Vision-Language Models: Attacks, Defenses and Evaluations
  7. 2025Eva-VLA: Evaluating Vision-Language-Action Models' Robustness Under Real-World Physical Variations
  8. 2026LIBERO-Safety: A Comprehensive Benchmark for Physical and Semantic Safety in Vision-Language-Action Models