Adnan KhanAdnanProduct & AI
San Francisco Bay Area
Get in touch

Source note3 min read

Technical Report on the CVPR 2026@AdvML Workshop Challenge

Source

arXiv

arxiv.org/abs/2607.11560v1

Read the source

Source

  • Source name: arXiv
  • Source kind: paper
  • Authors / organization: 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, Zeyang Zhang, Xiang Zhang, Yongxuan Zhu, Rui Ding, Hainan Li, Yongkang Zhang, Zhilei Zhu, Xianglong Kong, Jin Hu, Zonghao Ying, Yisong Xiao, Lei Chen, Haotong Qin, Jiakai Wang, Aishan Liu, Ruikai Li, Julia Karbing, Yinpeng Dong, Zhenfei Yin, Shao Jing, Xia Hu, Jingyi Xu, Juntao Dai, Xinyun Chen, Vishal M. Patel, Xianglong Liu, Dawn Song, Alan Yuille, Philip H. S. Torr, Dacheng Tao
  • Venue: arXiv: cs.CV
  • Published: 2026-07-13
  • Collected: 2026-07-14
  • URL: http://arxiv.org/abs/2607.11560v1
  • PDF: https://arxiv.org/pdf/2607.11560v1
  • DOI: UNKNOWN

Plain-Language Brief

This text is about the CVPR 2026@AdvML Workshop Challenge, focusing on adversarial attacks against Vision-Language Agents (VLAs) used in autonomous driving. The challenge involves creating adversarial images and textual perturbations to mislead models while preserving image quality and minimizing textual changes. The competition has two phases, with the second phase including a hidden black-box model for assessing transferability. The report discusses the task design, submission rules, evaluation protocol, and leaderboard results, as well as analyzing five leading submissions.

What the Document Says

  • The CVPR 2026@AdvML Workshop Challenge is focused on adversarial multimodal attacks against autonomous-driving Vision-Language Agents (VLAs).
  • The challenge uses DriveLM-style multi-view visual question answering to represent scenes with six synchronized camera images and driving-related question-answer pairs.
  • Participants create adversarial images and textual perturbations that induce model responses to deviate from reference answers while preserving image fidelity and limiting textual cost.
  • The competition consists of two phases, with Phase II adding a hidden black-box model for assessing transferability.
  • Several recurring patterns emerged in technical reports of leading submissions, such as favoring image-side attacks due to the suffix penalty, using scene-level, multi-view optimization over treating views in isolation, utilizing QA types and graph structure to allocate attack budget, improving black-box transfer with feature-space objectives, and finding a persistent vulnerability in driving VLAs from typographic content embedded in camera images.

Key Concepts Mentioned

  • Vision-Language Agents
  • Autonomous driving
  • Adversarial attacks
  • CVPR 2026@AdvML Workshop Challenge
  • DriveLM-style multi-view visual question answering
  • Image fidelity
  • Textual cost
  • Leaderboard results
  • Technical reports
  • Black-box model
  • Transferability

Claims Made by the Document

  • Vision-Language Agents are increasingly used in complex driving scenes and safety-critical reasoning.
  • The CVPR 2026@AdvML Workshop Challenge was designed to evaluate the robustness of autonomous-driving Vision-Language Agents against adversarial attacks.
  • Image-side attacks are favored by the suffix penalty in the challenge.

Evidence Used

  • Increasing use of VLAs for complex driving scenes and safety-critical reasoning is implied throughout the text.
  • The description of CVPR 2026@AdvML Workshop Challenge explicitly states its focus on adversarial multimodal attacks against autonomous-driving VLAs.
  • The mention of suffix penalty suggests a preference for image-side attacks.

Methods / Approach

UNKNOWN

Limitations Stated by the Authors

UNKNOWN

Limitations Noted by the Agent

Agent interpretation — not from the source.

UNKNOWN

Open Questions

UNKNOWN

Notes for Future Q&A

May include interpretation. Labeled where it does.

UNKNOWN

Raw Abstract / Excerpt

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 multi-view visual question answering, the challenge represents each scene with six synchronized camera images and a structured collection of driving-related question-answer pairs. Participants generate adversarial images and suffix-only textual perturbations that induce model responses to deviate from reference answers while preserving image fidelity and limiting textual cost. The competition comprises two phases, with Phase II adding a hidden black-box model to assess transferability. We describe the task design, submission rules, evaluation protocol, and leaderboard results, and then examine five leading submissions for which technical reports were available. Across these reports, several recurring patterns emerge: image-side attacks are favored by the suffix penalty; scene-level, multi-view optimization is more effective than treating views in isolation; QA types and graph structure provide useful priors for allocating attack budget; feature-space objectives can improve black-box transfer; and typographic content embedded in camera images exposes a persistent vulnerability in driving VLAs. These findings provide a practical reference for future robustness evaluation and defense design in multimodal autonomous-driving systems.