English(EN) Feature-Aware Token Attack for Compression-Triggered Stealthy Failures in Large Vision-Language Models

新攻击利用压缩引发的视觉语言模型故障

作者 PulseAugur 编辑部 · [1 个来源] · 2026-10-01 04:00

研究人员开发了一种名为特征感知令牌攻击（FATA）的新型对抗性攻击方法，旨在利用大型视觉语言模型（VLMs）中因视觉令牌压缩而产生的漏洞。该攻击旨在导致隐蔽故障，即模型在处理完整令牌时表现正常，但在压缩后出现错误，即使两个推理路径在初始干净图像上都能成功。FATA通过抑制注意力同时保留显著令牌的基于余弦的特征来实现这一点，在各种压缩器和任务上对LLaVA-1.5-7B模型展示了高准确度保持和条件致盲。 AI

影响这项研究突显了压缩VLMs中潜在的安全漏洞，需要进一步研究鲁棒的评估方法。

排序理由该集群包含一篇详细介绍大型视觉语言模型新型攻击方法的学术论文。[lever_c_demoted from research: ic=1 ai=1.0]

在 arXiv cs.CV 阅读 →

AI 生成摘要 · Google Gemini · 来自 1 个来源。我们如何撰写摘要 →

本文如何被排名

Signal score: 21 / 100; Composite score across the factors below. Higher = stronger signal that this story matters right now.
Newsworthiness bucket: Tool; 该集群包含一篇详细介绍大型视觉语言模型新型攻击方法的学术论文。[lever_c_demoted from research: ic=1 ai=1.0]
Source corroboration: Single-source cluster; Only one publisher covered this so far. Single-source stories can still rank when the publisher is high-authority, but they lack cross-source corroboration.
Topics: paper, safety; Editorial topic classification. Feeds into how the story surfaces on /topic/<slug> hub pages and into the per-entity coverage mix.
AI-industry relevance: High; Clearly on-topic for AI-industry coverage.
Story freshness: Breaking (< 6h); Fresh story with cross-source coverage still developing. Ranking may shift as more sources report.

完整方法见我们的编辑标准。

报道来源 [1]

arXiv cs.CV TIER_1 English(EN) · Shilinlu Yan, Bowen Chen, Yuechen Zhang, Zhenhong Zhou, Li Sun, Sen Su · 2026-10-01 04:00

面向压缩触发的大型视觉语言模型隐蔽故障的特征感知令牌攻击

arXiv:2609.39134v1 Announce Type: new Abstract: Visual-token compression improves the efficiency of large vision-language models, but can expose failures that full-token evaluation misses. We study adversarial images that preserve full-token correctness yet induce errors after co…

本文如何被排名

报道来源 [1]

面向压缩触发的大型视觉语言模型隐蔽故障的特征感知令牌攻击

相关实体

相关话题