Are Vision-Language-Action Models Robust to One-Step Observation Perturbations?
cs.AI, cs.LG, cs.RO
Submitted: 2026-09-26
Updated: 2026-09-26
Code: https://github.com/huggingface/lerobot
License: http://creativecommons.org/licenses/by-nc-nd/4.0/
The gist: Understanding the safety risks of vision-language-action (VLA) models is essential for their deployment in the physical world.
Terminology
Abstract
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.
Sources
- $\pi_0$: A Vision-Language-Action Flow Model for General Robot Control
- Mixture of Horizons in Action Chunking
- Adversarial Attacks on Robotic Vision Language Action Models
- CronusVLA: Towards Efficient and Robust Manipulation via Multi-Frame Vision-Language-Action Modeling
- Vision-Language-Action Safety: Threats, Challenges, Evaluations, and Mechanisms
- VLA-Corrector: Lightweight Detect-and-Correct Inference for Adaptive Action Horizon
- FAST: Efficient Action Tokenization for Vision-Language-Action Models
- SmolVLA: A Vision-Language-Action Model for Affordable and Efficient Robotics
- VLA Knows Its Limits: Adaptive Execution Horizons for Robot Policies
- When to Trust Imagination: Adaptive Action Execution for World Action Models
- STRONG-VLA: Decoupled Robustness Learning for Vision-Language-Action Models under Multimodal Perturbations
- Model-agnostic Adversarial Attack and Defense for Vision-Language-Action Models
- RobustVLA: Robustness-Aware Reinforcement Post-Training for Vision-Language-Action Models
- Learning Fine-Grained Bimanual Manipulation with Low-Cost Hardware
Related papers
- MAVEN-T: Reinforced Heterogeneous Distillation for Real-Time Multi-Agent Trajectory Prediction
- Model Discovery Agent: LLM-assisted Bayesian experiment design for data-efficient discovery of mechanistic world models
- Agents' Last Exam
- The Clinician's Veto: Navigating Trust, Liability, and Uncertainty in Autonomous AI Prescribing
- MindHelper: Closed-Loop Embodied Mental-State Reasoning for Precision Intervention
- Incumbent Advantage: Brand Bias and Cognitive Manipulation Dynamics in LLM Recommendation Systems