Saved in:
Bibliographic Details
Main Authors: Choi, Jinho, Lim, Hyesu, Schneider, Steffen, Choo, Jaegul
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2510.26186
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866917050223427584
author Choi, Jinho
Lim, Hyesu
Schneider, Steffen
Choo, Jaegul
author_facet Choi, Jinho
Lim, Hyesu
Schneider, Steffen
Choo, Jaegul
contents Dataset bias, where data points are skewed to certain concepts, is ubiquitous in machine learning datasets. Yet, systematically identifying these biases is challenging without costly, fine-grained attribute annotations. We present ConceptScope, a scalable and automated framework for analyzing visual datasets by discovering and quantifying human-interpretable concepts using Sparse Autoencoders trained on representations from vision foundation models. ConceptScope categorizes concepts into target, context, and bias types based on their semantic relevance and statistical correlation to class labels, enabling class-level dataset characterization, bias identification, and robustness evaluation through concept-based subgrouping. We validate that ConceptScope captures a wide range of visual concepts, including objects, textures, backgrounds, facial attributes, emotions, and actions, through comparisons with annotated datasets. Furthermore, we show that concept activations produce spatial attributions that align with semantically meaningful image regions. ConceptScope reliably detects known biases (e.g., background bias in Waterbirds) and uncovers previously unannotated ones (e.g, co-occurring objects in ImageNet), offering a practical tool for dataset auditing and model diagnostics.
format Preprint
id arxiv_https___arxiv_org_abs_2510_26186
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle ConceptScope: Characterizing Dataset Bias via Disentangled Visual Concepts
Choi, Jinho
Lim, Hyesu
Schneider, Steffen
Choo, Jaegul
Computer Vision and Pattern Recognition
Artificial Intelligence
Dataset bias, where data points are skewed to certain concepts, is ubiquitous in machine learning datasets. Yet, systematically identifying these biases is challenging without costly, fine-grained attribute annotations. We present ConceptScope, a scalable and automated framework for analyzing visual datasets by discovering and quantifying human-interpretable concepts using Sparse Autoencoders trained on representations from vision foundation models. ConceptScope categorizes concepts into target, context, and bias types based on their semantic relevance and statistical correlation to class labels, enabling class-level dataset characterization, bias identification, and robustness evaluation through concept-based subgrouping. We validate that ConceptScope captures a wide range of visual concepts, including objects, textures, backgrounds, facial attributes, emotions, and actions, through comparisons with annotated datasets. Furthermore, we show that concept activations produce spatial attributions that align with semantically meaningful image regions. ConceptScope reliably detects known biases (e.g., background bias in Waterbirds) and uncovers previously unannotated ones (e.g, co-occurring objects in ImageNet), offering a practical tool for dataset auditing and model diagnostics.
title ConceptScope: Characterizing Dataset Bias via Disentangled Visual Concepts
topic Computer Vision and Pattern Recognition
Artificial Intelligence
url https://arxiv.org/abs/2510.26186