Metraipolis › archive › dc_4abea5e775be

Can't Find Waldo: Evaluating VLMs' Sensitivity to Image Resolution and Detail Level — arXiv:2609.31706

Source
arXiv (en)
Original
https://arxiv.org/abs/2609.31706
License (as archived)
CC0-1.0
Archived
2026-10-07T20:59:25.416Z
SHA-256 of the archived text
d07081e1160812b203eced7c422ea9cdeaaa8df2fe153d0061dc109e74d393f3
Stable URL of this page
https://metraipolis.com/documento/dc_4abea5e775be

How to cite

Can't Find Waldo: Evaluating VLMs' Sensitivity to Image Resolution and Detail Level — arXiv:2609.31706 — arXiv (en), https://arxiv.org/abs/2609.31706 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_4abea5e775be (Metraipolis archive dc_4abea5e775be, sha256 d07081e1160812b203eced7c422ea9cdeaaa8df2fe153d0061dc109e74d393f3)
BibTeX
@misc{dc4abea5e775be,
  title = {Can't Find Waldo: Evaluating VLMs' Sensitivity to Image Resolution and Detail Level — arXiv:2609.31706},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2609.31706},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_4abea5e775be; sha256 d07081e1160812b203eced7c422ea9cdeaaa8df2fe153d0061dc109e74d393f3},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.31706. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Visual Language Models (VLMs) have achieved remarkable success across diverse tasks, yet they struggle with high-resolution inputs where critical information resides in small regions or detailed, cluttered scenes. While several approaches address this limitation, a systematic understanding of why models fail at high resolutions is lacking. We introduce a controlled evaluation framework that disentangles resolution-related performance degradation from task difficulty through semantics-preserving transformations. We propose two simple metrics: Area Under the Scaling Curve (AUSC), which quantifies scaling robustness independent of baseline accuracy, and Prediction Variance Score (PVS), which measures resolution-induced prediction instability. Through comprehensive experiments across 5 model families and 5 benchmarks, we identify three primary failure modes: (1) information loss from downsampling at vision token limits, (2) tokenization artifacts from patch boundary shifts and positional encoding fragility under non-standard aspect ratios, and (3) attention dilution as token counts increase. Our analysis reveals that even state-of-the-art models suffer from performance drops when processing high-resolution images, with degradation patterns varying systematically by architectural family. We provide actionable insights for model architecture design and data augmentation strategies to mitigate these limitations.

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-29 in cs.AI).
Categorie arXiv: Computer Vision and Pattern Recognition (cs.CV) ; Artificial Intelligence (cs.AI).
Pagina del lavoro: https://arxiv.org/abs/2609.31706 — 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 2026-10-07T20:29:02.037Z dalla pagina https://arxiv.org/catchup/cs.AI/2026-09-29?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).