mlsanityData to 14 Oct 2026

ReasonBreak: Probing Vulnerabilities in Reasoning-Enabled Vision-Language-Action Models for Autonomous Driving

arXiv 2605.29114 ↗First posted 27 May 2026cs.CR cs.LG cs.ROcited by 0

On the map: Vision-language-action models → VLA for autonomous driving

also in 1 more
Abstract (arXiv)

Vision-Language-Action (VLA) models with integrated reasoning have been proposed for end-to-end autonomous driving, assuming a tight coupling between reasoning and trajectory generation. However, the robustness of such systems under realistic input perturbations remains largely unexplored. We show that these models are highly vulnerable to realistic input perturbations, achieving up to 89% attack success rate (ASR) on reasoning and up to 72% on trajectory manipulation in closed-loop simulation, leading to increased collision rates and degraded safety metrics. Using NVIDIA's recent Alpamayo models as representative industry-developed VLAs, we conduct the first systematic black-box study of reasoning-enabled VLA models under realistic textual input corruptions, evaluating their impact on reasoning and driving behavior. We introduce a reasoning-aware evaluation framework capturing both semantic and structural aspects of reasoning, along with safety-centric measures. We also introduce a benchmark for evaluating attacks and defenses on reasoning-trajectory interactions in autonomous driving. Our results highlight the need for rigorous evaluation and improved defenses to ensure the safety of reasoning-enabled VLA systems in autonomous driving.

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Placed in VLA for autonomous driving, one of 13 directions in Vision-language-action models. Builds on 42 of its 55 references that are in the map. First posted on arXiv 27 May 2026.

Builds on

42 of 55 references.

Closest ancestors · 5

Same direction, newest first

  1. 2026ICR-Drive: Instruction Counterfactual Robustness for End-to-End Language-Driven Autonomous Driving7 Apr
  2. 2026HiST-VLA: A Hierarchical Spatio-Temporal Vision-Language-Action Model for End-to-End Autonomous Driving11 Feb
  3. 2025Alpamayo-R1: Bridging Reasoning and Action Prediction for Generalizable Autonomous Driving in the Long Tail30 Oct
  4. 2025AutoVLA: A Vision-Language-Action Model for End-to-End Autonomous Driving with Adaptive Reasoning and Reinforcement Fine-Tuning16 Jun
  5. 2025OpenDriveVLA: Towards End-to-end Autonomous Driving with Large Vision Language Action Model30 Mar

Other references · 37

  1. 2024OpenVLA: An Open-Source Vision-Language-Action Modelcited by 2,887
  2. 2017CARLA: An Open Urban Driving Simulatorcited by 2,449
  3. 2024π0: A Vision-Language-Action Flow Model for General Robot Controlcited by 2,407
  4. 2016End to End Learning for Self-Driving Carscited by 1,009
  5. 2025SpatialVLA: Exploring Spatial Representations for Visual-Language-Action Modelcited by 519
  6. 2017End-to-end Driving via Conditional Imitation Learningcited by 485
  7. 2022TransFuser: Imitation with Transformer-Based Sensor Fusion for Autonomous Drivingcited by 485
  8. 2019Exploring the Limitations of Behavior Cloning for Autonomous Drivingcited by 346
  9. 2024TinyVLA: Towards Fast, Data-Efficient Vision-Language-Action Models for Robotic Manipulationcited by 339
  10. 2025MolmoAct: Action Reasoning Models that can Reason in Spacecited by 197
  11. 2025HybridVLA: Collaborative Diffusion and Autoregression in a Unified Vision-Language-Action Modelcited by 181
  12. 2025GraspVLA: a Grasping Foundation Model Pre-trained on Billion-scale Synthetic Action Datacited by 123
25 more
  1. 2025OverThink: Slowdown Attacks on Reasoning LLMscited by 96
  2. 2024Best-of-N Jailbreakingcited by 76
  3. 2024Exploring the Adversarial Vulnerabilities of Vision-Language-Action Models in Roboticscited by 73
  4. 2025InstructVLA: Vision-Language-Action Instruction Tuning from Understanding to Manipulationcited by 72
  5. 2025BadVLA: Towards Backdoor Attacks on Vision-Language-Action Models via Objective-Decoupled Optimizationcited by 38
  6. 2024Visual Adversarial Attack on Vision-Language Models for Autonomous Drivingcited by 36
  7. 2025Adversarial Attacks on Robotic Vision Language Action Modelscited by 32
  8. 2024Reasoning Robustness of LLMs to Adversarial Typographical Errorscited by 31
  9. 2025Black-Box Adversarial Attack on Vision Language Models for Autonomous Drivingcited by 24
  10. 2025Model-agnostic Adversarial Attack and Defense for Vision-Language-Action Modelscited by 22
  11. 2024Speech Robust Bench: A Robustness Benchmark For Speech Recognitioncited by 21
  12. 2025FreezeVLA: Action-Freezing Attacks against Vision-Language-Action Modelscited by 21
  13. 2025When Alignment Fails: Multimodal Adversarial Attacks on Vision-Language-Action Modelscited by 20
  14. 2025Excessive Reasoning Attack on Reasoning LLMscited by 18
  15. 2025When Robots Obey the Patch: Universal Transferable Patch Attacks on Vision-Language-Action Modelscited by 13
  16. 2025Attention-Guided Patch-Wise Sparse Adversarial Attacks on Vision-Language-Action Modelscited by 11
  17. 2022CI-AVSR: A Cantonese Audio-Visual Speech Dataset for In-car Command Recognitioncited by 10
  18. 2025Evaluating Robustness of Large Language Models Against Multilingual Typographical Errorscited by 9
  19. 2025Phantom Menace: Exploring and Enhancing the Robustness of VLA Models Against Physical Sensor Attackscited by 9
  20. 2021An Approach to Improve Robustness of NLP Systems against ASR Errorscited by 8
  21. 2025Practical Reasoning Interruption Attacks on Reasoning Large Language Modelscited by 7
  22. 2026ReasoningBomb: A Stealthy Denial-of-Service Attack by Inducing Pathologically Long Reasoning in Large Reasoning Modelscited by 5
  23. 2026TRAP: Hijacking VLA CoT-Reasoning via Adversarial Patchescited by 5
  24. 2024Decictor: Towards Evaluating the Robustness of Decision-Making in Autonomous Driving Systemscited by 4
  25. 2026SABER: A Stealthy Agentic Black-Box Attack Framework for Vision-Language-Action Modelscited by 2

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2026Is VLA Reasoning Faithful? Probing Safety of Chain-of-Causation in Autonomous Driving Models
  2. 2025Reasoning-VLA: A Fast and General Vision-Language-Action Reasoning Model for Autonomous Driving
  3. 2025A Survey on Vision-Language-Action Models for Autonomous Driving
  4. 2026SafeDriveVLA: Navigation-Conditioned World Model Dreaming for Conflict-Aware End-to-End Autonomous Driving
  5. 2026RT-VLA: Real-Time Vision-Language-Action Models via Knowledge Distillation
  6. 2026Lost in Fog: Sensor Perturbations Expose Reasoning Fragility in Driving VLAs
  7. 2025Counterfactual VLA: Self-Reflective Vision-Language-Action Model with Adaptive Reasoning
  8. 2026Drive the Thoughts: Runtime Monitoring of VLA Reasoning-Trajectory Consistency