Person-Centric Annotations of LAION-400M: Auditing Bias and Its Transfer to Models

Leander Girrbach, Stephan Alaniz, Genevieve Smith, Trevor Darrell, Zeynep Akata
10/4/2025
cs.CVcs.CLcs.CY

No official code found. Best implementation: ExplainableML/LAION-400M-Person-Centric-Annotations (3 stars, last commit 2026-02-13).

Abstract

Vision-language models trained on large-scale multimodal datasets show strong demographic biases, but the role of training data in producing these biases remains unclear. A major barrier has been the lack of demographic annotations in web-scale datasets such as LAION-400M. We address this gap by creating person-centric annotations for the full dataset, including over 276 million bounding boxes, perceived gender and race/ethnicity labels, and automatically generated captions. These annotations are produced through validated automatic labeling pipelines combining object detection, multimodal captioning, and finetuned classifiers. Using them, we uncover demographic imbalances and harmful associations, such as the disproportionate linking of men and individuals perceived as Black or Middle Eastern with crime-related and negative content. We also show that 60-70% of gender bias in CLIP and Stable Diffusion can be linearly explained by direct co-occurrences in the data. Our resources establish the first large-scale empirical link between dataset composition and downstream model bias.

View on arXivView PDF

Code Implementations(1)

ExplainableML/LAION-400M-Person-Centric-AnnotationsCommunity74%

[ICLR 2026] Official codebase of "Person-Centric Annotations of LAION-400M: Auditing Bias and Its Transfer to Models"

30Python, Jupyter NotebookFeb 13, 20268 months ago
automatic-annotationdataset-biasdemographic-biaslaion400msocial-bias

Similar Papers

Data Matters Most: Auditing Social Bias in Contrastive Vision Language Models

85% match2025

Assessing the Reliability of LLMs Annotations in the Context of Demographic Bias and Model Explanation

85% match2025

Uncurated Image-Text Datasets: Shedding Light on Demographic Bias

84% match1 repo182026

Robustness and Confounders in the Demographic Alignment of LLMs with Human Perceptions of Offensiveness

84% match1 repo2024

Evaluating LLMs for Demographic-Targeted Social Bias Detection: A Comprehensive Benchmark Study

84% match2025

A Unified Framework and Dataset for Assessing Societal Bias in Vision-Language Models

84% match1 repo12024

Cite this paper

@article{girrbach2025personcentric,
  title  = {Person-Centric Annotations of LAION-400M: Auditing Bias and Its Transfer to Models},
  author = {Leander Girrbach and Stephan Alaniz and Genevieve Smith and Trevor Darrell and Zeynep Akata},
  year   = {2025},
  eprint = {2510.03721},
  archivePrefix = {arXiv},
  url    = {http://arxiv.org/abs/2510.03721v1}
}

Add a badge to your README

Shows whether official code is linked for this paper.

Code of Paper badge
[![Code of Paper](https://codeofpaper.com/badge/2510.03721.svg)](https://codeofpaper.com/paper/2510.03721)

Discussion