Metraipolis › archive › dc_e8edef1f311d
Sorry Robot, Happy Human: Vision-Language Models Read Only One of Two Legible Typographic Layers — arXiv:2609.31403
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2609.31403
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-08T01:39:56.133Z
- SHA-256 of the archived text
62472e6794e1ad4bc294275b3fba3f2f985f9d09e25892aac942f92c265cddae- Stable URL of this page
https://metraipolis.com/documento/dc_e8edef1f311d
How to cite
Sorry Robot, Happy Human: Vision-Language Models Read Only One of Two Legible Typographic Layers — arXiv:2609.31403 — arXiv (en), https://arxiv.org/abs/2609.31403 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_e8edef1f311d (Metraipolis archive dc_e8edef1f311d, sha256 62472e6794e1ad4bc294275b3fba3f2f985f9d09e25892aac942f92c265cddae)
BibTeX
@misc{dce8edef1f311d,
title = {Sorry Robot, Happy Human: Vision-Language Models Read Only One of Two Legible Typographic Layers — arXiv:2609.31403},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2609.31403},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_e8edef1f311d; sha256 62472e6794e1ad4bc294275b3fba3f2f985f9d09e25892aac942f92c265cddae},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.31403. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
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. Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-28 in cs.AI). Categorie arXiv: Computation and Language (cs.CL) ; Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV). Pagina del lavoro: https://arxiv.org/abs/2609.31403 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito). Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina. Letto il 2026-10-08T01:12:14.592Z dalla pagina https://arxiv.org/catchup/cs.AI/2026-09-28?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).