研究人员开发了FRA-Attack，一种提高多模态大语言模型（MLLMs）对抗性攻击迁移能力的新方法。该技术利用频域正则化将扰动与不同模型共享的视觉线索对齐，克服了现有空间域方法的局限性。在15个MLLMs上的实验表明，FRA-Attack的性能优越，特别是针对GPT-5.4、Claude-Opus-4.6和Gemini-3-flash等模型。 AI
影响 增强了对MLLM漏洞的理解，并为安全研究提供了信息。
排序理由 该集群包含一篇详细介绍新研究方法的学术论文。
在 Hugging Face Daily Papers 阅读 →
AI 生成摘要 · Google Gemini · 来自 3 个来源。 我们如何撰写摘要 →
Multimodal large language models (MLLMs) remain vulnerable to transfer-based targeted attacks, where perturbations optimized on open-source surrogate encoders can generalize to closed-source MLLMs. A key challenge for improving adversarial transferability is to effectively captur…
arXiv:2605.21541v1 Announce Type: cross Abstract: Multimodal large language models (MLLMs) remain vulnerable to transfer-based targeted attacks, where perturbations optimized on open-source surrogate encoders can generalize to closed-source MLLMs. A key challenge for improving ad…
Multimodal large language models (MLLMs) remain vulnerable to transfer-based targeted attacks, where perturbations optimized on open-source surrogate encoders can generalize to closed-source MLLMs. A key challenge for improving adversarial transferability is to effectively captur…