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

arXiv: https://arxiv.org/abs/2501.13223 · HTML: https://mlsanity.com/paper/2501.13223/

## Key facts

- Cited by 1 papers in the map (data through 14 Oct 2026).
- Placed in Social bias and fairness in VLMs, one of 16 directions in Multimodal LLMs / vision-language models.
- Builds on 21 of its 28 references that are in the map.
- First posted on arXiv 22 Jan 2025.

## Placed in

- Multimodal LLMs / vision-language models / [Social bias and fairness in VLMs](https://mlsanity.com/mllm/social-bias-and-fairness-in-vlms/)

## Builds on (most cited first, 21 of 21)

- [Learning Transferable Visual Models From Natural Language Supervision](https://mlsanity.com/paper/2103.00020/) (2021)
- [Training Compute-Optimal Large Language Models](https://mlsanity.com/paper/2203.15556/) (2022)
- [Reproducible scaling laws for contrastive language-image learning](https://mlsanity.com/paper/2212.07143/) (2022)
- [Multimodal datasets: misogyny, pornography, and malignant stereotypes](https://mlsanity.com/paper/2110.01963/) (2021)
- [Towards Understanding and Mitigating Social Biases in Language Models](https://mlsanity.com/paper/2106.13219/) (2021)
- [Debiasing Vision-Language Models via Biased Prompts](https://mlsanity.com/paper/2302.00070/) (2023)
- [Evaluating CLIP: Towards Characterization of Broader Capabilities and Downstream Implications](https://mlsanity.com/paper/2108.02818/) (2021)
- [A Prompt Array Keeps the Bias Away: Debiasing Vision-Language Models with Adversarial Learning](https://mlsanity.com/paper/2203.11933/) (2022)
- [DeAR: Debiasing Vision-Language Models with Additive Residuals](https://mlsanity.com/paper/2303.10431/) (2023)
- [VHELM: A Holistic Evaluation of Vision Language Models](https://mlsanity.com/paper/2410.07112/) (2024)
- [VisoGender: A dataset for benchmarking gender bias in image-text pronoun resolution](https://mlsanity.com/paper/2306.12424/) (2023)
- [On Hate Scaling Laws For Data-Swamps](https://mlsanity.com/paper/2306.13141/) (2023)
- [Identifying Implicit Social Biases in Vision-Language Models](https://mlsanity.com/paper/2411.00997/) (2024)
- [MultiModal Bias: Introducing a Framework for Stereotypical Bias Assessment beyond Gender and Race in Vision Language Models](https://mlsanity.com/paper/2303.12734/) (2023)
- [textttModSCAN: Measuring Stereotypical Bias in Large Vision-Language Models from Vision and Language Modalities](https://mlsanity.com/paper/2410.06967/) (2024)
- [SANER: Annotation-free Societal Attribute Neutralizer for Debiasing CLIP](https://mlsanity.com/paper/2408.10202/) (2024)
- [BendVLM: Test-Time Debiasing of Vision-Language Embeddings](https://mlsanity.com/paper/2411.04420/) (2024)
- [No Filter: Cultural and Socioeconomic Diversity in Contrastive Vision-Language Models](https://mlsanity.com/paper/2405.13777/) (2024)
- [BiasDora: Exploring Hidden Biased Associations in Vision-Language Models](https://mlsanity.com/paper/2407.02066/) (2024)
- [Social Perception of Faces in a Vision-Language Model](https://mlsanity.com/paper/2408.14435/) (2024)
- [FairCoT: Enhancing Fairness in Text-to-Image Generation via Chain of Thought Reasoning with Multimodal Large Language Models](https://mlsanity.com/paper/2406.09070/) (2024)

## Built on by (most cited first, 1 of 1)

- [On Fairness of Task Arithmetic: The Role of Task Vectors](https://mlsanity.com/paper/2505.24262/) (2025)

## Abstract (arXiv)

Vision-language models (VLMs) deliver strong zero-shot recognition but frequently inherit social biases from their training data. We systematically disentangle three design factors -- model size, training-data scale, and training-data source -- by comparing CLIP and OpenCLIP, two models that share an identical contrastive objective yet differ in encoder width and in the image-text corpora on which they are pre-trained (400M proprietary pairs vs. 400M/2B LAION). Across balanced face-analysis benchmarks, enlarging the encoder reduces gender skew in CLIP but amplifies both gender and racial skew in OpenCLIP; increasing the LAION corpus from 400M to 2B further increases OpenCLIP bias. At matched model and data budgets, substituting proprietary data with LAION improves gender fairness while increasing racial skew, underscoring data source as the primary driver of bias patterns. We also evaluate three post-hoc, test-time debiasing strategies -- Bias Prompts, Prompt Array, and SANER. Debiasing reduces but does not eliminate harm, and its effectiveness is source- and size-dependent: Bias Prompts most effectively reduce gender skew in CLIP at smaller model sizes, whereas Prompt Array and SANER more reliably reduce racial skew in OpenCLIP; scaling LAION reconfigures which method is most fair. Taken together, these findings challenge the assumption that bigger models or datasets are automatically fairer and foreground training data source as the key determinant of both bias and mitigation efficacy. We release code and evaluation scripts to enable transparent, reproducible auditing of future VLMs.