Saved in:
Bibliographic Details
Main Authors: Hrynenko, Olena, Baranouskaya, Darya, Baia, Alina Elena, Cavallaro, Andrea
Format: Preprint
Published: 2026
Subjects:
Online Access:https://arxiv.org/abs/2602.07931
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866915784673984512
author Hrynenko, Olena
Baranouskaya, Darya
Baia, Alina Elena
Cavallaro, Andrea
author_facet Hrynenko, Olena
Baranouskaya, Darya
Baia, Alina Elena
Cavallaro, Andrea
contents Visual Language Models (VLMs) are often used for zero-shot detection of visual attributes in the image. We present a zero-shot evaluation of open-source VLMs for privacy-related attribute recognition. We identify the attributes for which VLMs exhibit strong inter-annotator agreement, and discuss the disagreement cases of human and VLM annotations. Our results show that when evaluated against human annotations, VLMs tend to predict the presence of privacy attributes more often than human annotators. In addition to this, we find that in cases of high inter-annotator agreement between VLMs, they can complement human annotation by identifying attributes overlooked by human annotators. This highlights the potential of VLMs to support privacy annotations in large-scale image datasets.
format Preprint
id arxiv_https___arxiv_org_abs_2602_07931
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Which private attributes do VLMs agree on and predict well?
Hrynenko, Olena
Baranouskaya, Darya
Baia, Alina Elena
Cavallaro, Andrea
Computer Vision and Pattern Recognition
Visual Language Models (VLMs) are often used for zero-shot detection of visual attributes in the image. We present a zero-shot evaluation of open-source VLMs for privacy-related attribute recognition. We identify the attributes for which VLMs exhibit strong inter-annotator agreement, and discuss the disagreement cases of human and VLM annotations. Our results show that when evaluated against human annotations, VLMs tend to predict the presence of privacy attributes more often than human annotators. In addition to this, we find that in cases of high inter-annotator agreement between VLMs, they can complement human annotation by identifying attributes overlooked by human annotators. This highlights the potential of VLMs to support privacy annotations in large-scale image datasets.
title Which private attributes do VLMs agree on and predict well?
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2602.07931