Skip to content

Translation in progress Some entries still include text in the original language.

Failure-First·Indexed by this site · Originally published

论文提出面向 VLA 模型的注意力引导逐块稀疏对抗攻击

[Daily Paper] Attention-Guided Patch-Wise Sparse Adversarial Attacks on Vision-Language-Action Models

AI summary

一篇论文提出面向视觉-语言-动作(VLA)模型的注意力引导逐块稀疏对抗攻击,以稀疏的图像块扰动形式实施,并由注意力信息引导。

Read originalfailurefirst.org