RedVLA: Physical Red Teaming for Vision-Language-Action Models

arXiv:2604.22591v1 Announce Type: new Abstract: The real-world deployment of Vision-Language-Action (VLA) models remains limited by the risk of unpredictable and irreversible physical harm. However, we currently lack effective mechanisms to proactively detect these physical safety risks before deployment. To address this gap, we propose extbf{RedVLA}, the first red teaming framework for physical safety in VLA models. We systematically uncover unsafe behaviors through a two-stage process: (I) extbf{Risk Scenario Synthesis} constructs a valid and task-feasible initial risk scene. Specifically, it identifies critical interaction regions from benign trajectories and positions the risk factor within these regions, aiming to entangle it with the VLA's execution flow and elicit a target unsafe behavior. (II) extbf{Risk Amplification} ensures stable elicitation across heterogeneous models. It iteratively refines the risk factor state through gradient-free optimization guided by trajectory features. Experiments on six representative VLA models show that RedVLA uncovers diverse unsafe behaviors and achieves the ASR up to 95.5% within 10 optimization iterations. To mitigate these risks, we further propose SimpleVLA-Guard, a lightweight safety guard built from RedVLA-generated data. Our data, assets, and code are available href{https://redvla.github.io}{here}.

Related

• ViVa: A Video-Generative Value Model for Robot Reinforcement Learning
• Towards Deploying VLA without Fine-Tuning: Plug-and-Play Inference-Time VLA Policy Steering via Embodied Evolutionary Diffusion
• Long-Term Memory for VLA-based Agents in Open-World Task Execution
• Clutter-Robust Vision-Language-Action Models through Object-Centric and Geometry Grounding
• Reflection-Based Task Adaptation for Self-Improving VLA
• OmniVLA-RL: A Vision-Language-Action Model with Spatial Understanding and Online RL

Source: arXiv cs.RO | 2026-04-27