发表机构
Fırat University（菲拉特大学）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
本研究通过DecoyBench数据集评估六个闭源视觉语言模型，发现它们在高分辨率下仅能读取轮廓文本而忽略阴影文本，在低分辨率下则相反，表现出处理多空间频率排版层的固有局限。
AI 中文摘要
视觉语言模型（VLMs）尽管在光学字符识别（OCR）任务中取得了成功，但容易受到排版攻击的影响，并且对于包含多个文本层的图像具有脆弱的结构。在本研究中，使用Decoy Font方法创建了DecoyBench数据集。该数据集包含300张图像，每张图像包含具有清晰轮廓线的文本叠加在具有柔和阴影的另一个文本之上。使用该数据集，在两种不同的提示条件（朴素和引导）以及两种不同的分辨率（$512\ imes512$和$64\ imes64$）下，评估了来自三个不同模型系列的六个最近的闭源模型。一项验证研究表明，人类参与者能够以高准确率读取两个文本层。相比之下，这些模型，在大多数变体和两种提示方法下，在高分辨率下以接近人类的准确率读取轮廓文本，但几乎从未完全提取阴影文本。在低分辨率下，轮廓文本无法被模型或人类读取，而阴影文本可以以高准确率提取。研究结果表明，所评估的视觉语言模型在处理包含多个空间频率层的排版结构时表现出一致的行为限制。
英文摘要
Vision-language models (VLMs), despite their success in optical character recognition (OCR) tasks, are vulnerable to typographic attacks and have a fragile structure for images with multiple text layers. In this study, the DecoyBench dataset was created using the Decoy Font method. The dataset consists of 300 images, each containing text with sharp contour lines superimposed on another text with soft shading. Six recent closed-source models from three different model families were evaluated using this dataset under two different prompting conditions (naive and guided) and at two different resolutions ($512\times512$ and $64\times64$). A validation study showed that human participants could read both text layers with high accuracy. In contrast, the models, with most variants and both prompting methods, read the contour text with near-human accuracy at high resolution, but almost never fully extracted the shading text. At low resolution, the contour text could not be read by either the models or humans, while the shading text could be extracted with high accuracy. The findings indicate that the evaluated VLMs exhibit a consistent behavioral limitation when processing typographic structures containing multiple spatial frequency layers.
CommentsAccepted to the First Workshop on Document Intelligence and Understanding (DocInsights 2026), co-located with the 2026 Conference on Empirical Methods in Natural Language Processing (EMNLP 2026)
