All series
Series

Adversarial Robustness in VLA Models

2 articles · Last updated 2025-11-18

Contents

  1. 1
    Research

    RoboGCG

    Attackers can fully control a VLA-driven robot by appending just ~20 optimized text tokens to a normal instruction—no image manipulation, no model access at deployment.

  2. 2
    Research

    Model-agnostic Adversarial Attack and Defense

    A model-agnostic adversarial attack disrupts vision-language-action models by misaligning visual-text embeddings, while adversarial fine-tuning defends by learning perturbation-invariant representations.