mlsanityData to 7 Oct 2026

Lost in Fog: Sensor Perturbations Expose Reasoning Fragility in Driving VLAs

arXiv 2605.21446 ↗First posted 20 May 2026cs.RO cs.AIcited by 3

On the map: End-to-end autonomous driving → Reasoning VLA and VLM driving policies

Abstract (arXiv)

Interpretable autonomous driving planners depend not only on generating explanations, but also on those explanations remaining reliable under real-world sensor degradation. In this paper we present a controlled perturbation study of Vision-Language-Action (VLA) robustness in autonomous driving, evaluating Alpamayo R1 (10B parameters) across 1,996 scenarios under eight sensor perturbations (Gaussian noise at four intensities, two lighting extremes, and two fog levels; ~18,000 inference trials). We find that reasoning consistency is a high-fidelity indicator of trajectory reliability: when Chain-of-Causation (CoC) explanations change after perturbation, trajectory deviation spikes 5.3× (21.8m vs 4.1m), with r!=!0.99 across attack types and rpb!=!0.53 per-sample (Cohen's d!=!1.12). A controlled ablation provides evidence that enabling CoC generation is associated with improved trajectory accuracy (11.8% on average across conditions; p < 0.0001) under matched inference settings. Over the tested noise range (σ ∈ 10, 30, 50, 70), degradation is approximately linear (R2!=!0.957), while standard input preprocessing defenses provide only marginal relief. Together, these results establish CoC consistency as a quantitative proxy for planning safety and motivate reasoning-based runtime monitoring for safer VLA deployment.

Key facts. Cited by 3 papers in the map (data through 7 Oct 2026). Placed in Reasoning VLA and VLM driving policies, one of 9 directions in End-to-end autonomous driving. Builds on 24 of its 30 references that are in the map. First posted on arXiv 20 May 2026.

Builds on

24 of 30 references.

Closest ancestors · 1

Same direction, newest first

  1. 2025Alpamayo-R1: Bridging Reasoning and Action Prediction for Generalizable Autonomous Driving in the Long Tail30 Oct

Other references · 23

  1. 2019nuScenes: A multimodal dataset for autonomous drivingcited by 4,485
  2. 2023RT-2: Vision-Language-Action Models Transfer Web Knowledge to Robotic Controlcited by 2,993
  3. 2017CARLA: An Open Urban Driving Simulatorcited by 2,431
  4. 2024Octo: An Open-Source Generalist Robot Policycited by 1,427
  5. 2016End to End Learning for Self-Driving Carscited by 1,000
  6. 2021KITTI-360: A Novel Dataset and Benchmarks for Urban Scene Understanding in 2D and 3Dcited by 570
  7. 2023DriveLM: Driving with Graph Visual Question Answeringcited by 527
  8. 2017Semantic Foggy Scene Understanding with Synthetic Datacited by 401
  9. 2019Benchmarking Robustness in Object Detection: Autonomous Driving when Winter is Comingcited by 332
  10. 2018Textual Explanations for Self-Driving Vehiclescited by 226
  11. 2023Rethinking the Open-Loop Evaluation of End-to-End Autonomous Driving in nuScenescited by 191
  12. 2019Adversarial Sensor Attack on LiDAR-based Perception in Autonomous Drivingcited by 151
11 more
  1. 2020Physically Realizable Adversarial Examples for LiDAR Object Detectioncited by 137
  2. 2018Model Adaptation with Synthetic and Real Data for Semantic Dense Foggy Scene Understandingcited by 98
  3. 2021AugMax: Adversarial Composition of Random Augmentations for Robust Trainingcited by 89
  4. 2022On Adversarial Robustness of Trajectory Prediction for Autonomous Vehiclescited by 89
  5. 2021Explainable Artificial Intelligence for Autonomous Driving: A Comprehensive Overview and Field Guide for Future Research Directionscited by 73
  6. 2017Explaining How a Deep Neural Network Trained with End-to-End Learning Steers a Carcited by 65
  7. 2024Visual Adversarial Attack on Vision-Language Models for Autonomous Drivingcited by 35
  8. 2025Eva-VLA: Evaluating Vision-Language-Action Models' Robustness Under Real-World Physical Variationscited by 24
  9. 2025RobustVLA: On Robustness of Vision-Language-Action Model against Multi-Modal Perturbationscited by 23
  10. 2025RoboDriveVLM: A Novel Benchmark and Baseline towards Robust Vision-Language Models for Autonomous Drivingcited by 8
  11. 2025Runtime Safety Monitoring of Deep Neural Networks for Perception: A Surveycited by 2

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2026: 3.

3 papers.

  1. 2026Imagined Rollouts are Kinematic, Not Dynamic: A Diagnosis of Long-Horizon World-Model Failurecited by 1
  2. 2026Output Type Before Quality: A Standards-Derived XAI Admissibility Rubric for Autonomous-Driving Safetycited by 0
  3. 2026When and How Severely: Scenario-Specific Safety Envelopes for Driving VLAscited by 0

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. 2026Probing the Reliability of Driving VLMs: From Inconsistent Responses to Grounded Temporal Reasoning
  3. 2026Reasoning models do not yet follow their reasoning in autonomous driving: The KITScenes LongTail Dataset
  4. 2026ReasonBreak: Probing Vulnerabilities in Reasoning-Enabled Vision-Language-Action Models for Autonomous Driving
  5. 2026Reasoning-aware Speculative Decoding for Efficient Vision-Language-Action Models in Autonomous Driving
  6. 2025Are VLMs Ready for Autonomous Driving? An Empirical Study from the Reliability, Data, and Metric Perspectives
  7. 2026ICR-Drive: Instruction Counterfactual Robustness for End-to-End Language-Driven Autonomous Driving
  8. 2026Drive the Thoughts: Runtime Monitoring of VLA Reasoning-Trajectory Consistency