TRAP: Hijacking VLA CoT-Reasoning via Adversarial Patches

Zhengxian Huang, Wenjun Zhu, Haoxun Qiu, Xiaoyu Ji†, Wenyuan Xu

International Conference on Machine Learning (ICML) 2026 Poster

the first targeted behavior-hijacking adversarial attack against CoT-reasoning VLA models via adversarial patch.

[Arxiv] [Project]

POEX: Towards Policy Executable Jailbreak Attacks Against the LLM-based Robots

Xuancun Lu, Zhengxian Huang, Xinfeng Li, Chi Zhang, Xiaoyu Ji, Wenyuan Xu

Arxiv 2024

a red-teaming framework that induces harmful yet executable policy to jailbreak LLM-based robots.

[Arxiv] [Code]