不要留下空白:视觉-语言模型中的无幻觉物体遮蔽
Do Not Leave a Gap: Hallucination-Free Object Concealment in Vision-Language Models
AI总结:
本文提出一种背景一致的物体遮蔽攻击方法,通过重新编码视觉表示使目标物体与背景一致,避免幻觉产生,实验显示其能有效遮蔽目标物体并减少幻觉。
摘要:
视觉-语言模型(VLMs)最近在视觉理解和生成方面表现出色,但易受对抗性操纵。先前的物体隐藏攻击主要通过抑制或阻断区域特定表示,常导致语义间隙,从而引发幻觉,即模型生成合理但错误的对象。本文证明幻觉并非源于物体缺失,而是由此类抑制攻击引入的语义不连续性所致。我们提出了一种新的背景一致物体遮蔽攻击方法,通过重新编码视觉表示使其在统计和语义上与周围背景一致。关键在于保留token结构和注意力流,避免触发幻觉的表示空缺。我们提出了一种像素级优化框架,在多个transformer层上强制背景一致的重新编码,同时保持全局场景语义。在最先进的视觉-语言模型上进行的广泛实验表明,我们的方法能有效遮蔽目标物体,同时保留多达86%的非目标物体,并将基于注意力抑制的攻击的幻觉减少高达3倍。
英文摘要:
Vision-language models (VLMs) have recently shown remarkable capabilities in visual understanding and generation, but remain vulnerable to adversarial manipulations of visual content. Prior object-hiding attacks primarily rely on suppressing or blocking region-specific representations, often creating semantic gaps that inadvertently induce hallucination, where models invent plausible but incorrect objects. In this work, we demonstrate that hallucination arises not from object absence per se, but from semantic discontinuity introduced by such suppression-based attacks. We propose a new class of \emph{background-consistent object concealment} attacks, which hide target objects by re-encoding their visual representations to be statistically and semantically consistent with surrounding background regions. Crucially, our approach preserves token structure and attention flow, avoiding representational voids that trigger hallucination. We present a pixel-level optimization framework that enforces background-consistent re-encoding across multiple transformer layers while preserving global scene semantics. Extensive experiments on state-of-the-art vision-language models show that our method effectively conceals target objects while preserving up to $86\%$ of non-target objects and reducing grounded hallucination by up to $3\times$ compared to attention-suppression-based attacks.
机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。