PulseAugur
EN
LIVE 10:36:44

CVPR 2026 challenge targets adversarial attacks on autonomous driving VLAs

By PulseAugur Editorial · [2 sources] ·

A technical report details the CVPR 2026@AdvML Workshop Challenge, which focused on adversarial multimodal attacks against autonomous-driving vision-language agents (VLAs). The challenge involved generating perturbations to induce incorrect responses from VLAs interpreting driving scenes, using multi-view visual question answering. Analysis of leading submissions revealed that image-side attacks are effective, scene-level optimization outperforms isolated view processing, and typographic content within images presents a persistent vulnerability. AI

IMPACT Highlights vulnerabilities in autonomous driving AI systems, informing future robustness and defense strategies.

RANK_REASON Technical report detailing a workshop challenge and its findings.

Read on arXiv cs.AI →

AI-generated summary · Google Gemini · from 2 sources. How we write summaries →

CVPR 2026 challenge targets adversarial attacks on autonomous driving VLAs

COVERAGE [2]

  1. arXiv cs.AI TIER_1 English(EN) · Tianyuan Zhang, Zonglei Jing, Jiangfan Liu, Ligong Zhang, Ke Ma, Chengzhi Sun, Xiaohai Xu, Zhirui Zhang, Qianqian Xu, Qingming Huang, Hanyu Fang, Junhua Liu, Zheng Wang, Xiaoliang Liu, Yuanbo Li, Shuai Gui, Bin Wang, Menghe Zheng, Jing Nie, Hanyang Meng,… ·

    Technical Report on the CVPR 2026@AdvML Workshop Challenge

    arXiv:2607.11560v1 Announce Type: cross Abstract: Vision-language agents (VLAs) are increasingly used to interpret complex driving scenes and support safety-critical reasoning. This report presents the CVPR 2026@AdvML Workshop Challenge on adversarial multimodal attacks against a…

  2. arXiv cs.AI TIER_1 English(EN) · Dacheng Tao ·

    Technical Report on the CVPR 2026@AdvML Workshop Challenge

    Vision-language agents (VLAs) are increasingly used to interpret complex driving scenes and support safety-critical reasoning. This report presents the CVPR 2026@AdvML Workshop Challenge on adversarial multimodal attacks against autonomous-driving VLAs. Built on DriveLM-style mul…

RELATED ENTITIES

RELATED TOPICS