mlsanityData to 7 Oct 2026

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

arXiv 2402.13636 ↗First posted 21 Feb 2024cs.CV cs.CL cs.CYcited by 11

On the map: Multimodal LLMs / vision-language models → Social bias and fairness in VLMs

Abstract (arXiv)

Vision-language models (VLMs) have gained widespread adoption in both industry and academia. In this study, we propose a unified framework for systematically evaluating gender, race, and age biases in VLMs with respect to professions. Our evaluation encompasses all supported inference modes of the recent VLMs, including image-to-text, text-to-text, text-to-image, and image-to-image. Additionally, we propose an automated pipeline to generate high-quality synthetic datasets that intentionally conceal gender, race, and age information across different professional domains, both in generated text and images. The dataset includes action-based descriptions of each profession and serves as a benchmark for evaluating societal biases in vision-language models (VLMs). In our comparative analysis of widely used VLMs, we have identified that varying input-output modalities lead to discernible differences in bias magnitudes and directions. Additionally, we find that VLM models exhibit distinct biases across different bias attributes we investigated. We hope our work will help guide future progress in improving VLMs to learn socially unbiased representations. We will release our data and code.

Key facts. Cited by 11 papers in the map (data through 7 Oct 2026). Placed in Social bias and fairness in VLMs, one of 16 directions in Multimodal LLMs / vision-language models. Builds on 32 of its 46 references that are in the map. First posted on arXiv 21 Feb 2024.

Builds on

32 of 46 references.

Closest ancestors · 2

Same direction, newest first

  1. 2024Examining Gender and Racial Bias in Large Vision-Language Models Using a Novel Dataset of Parallel Images8 Feb
  2. 2023VisoGender: A dataset for benchmarking gender bias in image-text pronoun resolution21 Jun

Other references · 30

  1. 2023Llama 2: Open Foundation and Fine-Tuned Chat Modelscited by 12,865
  2. 2022Hierarchical Text-Conditional Image Generation with CLIP Latentscited by 6,539
  3. 2022Photorealistic Text-to-Image Diffusion Models with Deep Language Understandingcited by 6,322
  4. 2019BERTScore: Evaluating Text Generation with BERTcited by 5,462
  5. 2023Improved Baselines with Visual Instruction Tuningcited by 4,362
  6. 2023SDXL: Improving Latent Diffusion Models for High-Resolution Image Synthesiscited by 4,290
  7. 2020StereoSet: Measuring stereotypical bias in pretrained language modelscited by 911
  8. 2020CrowS-Pairs: A Challenge Dataset for Measuring Social Biases in Masked Language Modelscited by 703
  9. 2020Null It Out: Guarding Protected Attributes by Iterative Nullspace Projectioncited by 442
  10. 2018Women also Snowboard: Overcoming Bias in Captioning Modelscited by 236
  11. 2022DALL-Eval: Probing the Reasoning Skills and Social Biases of Text-to-Image Generation Modelscited by 230
  12. 2020Measuring and Reducing Gendered Correlations in Pre-trained Modelscited by 222
18 more
  1. 2019Identifying and Reducing Gender Bias in Word-Level Language Modelscited by 214
  2. 2023Any-to-Any Generation via Composable Diffusioncited by 203
  3. 2019A Framework for Understanding Sources of Harm throughout the Machine Learning Life Cyclecited by 192
  4. 2020Towards Debiasing Sentence Representationscited by 170
  5. 2023ViP-LLaVA: Making Large Multimodal Models Understand Arbitrary Visual Promptscited by 160
  6. 2017No Classification without Representation: Assessing Geodiversity Issues in Open Data Sets for the Developing Worldcited by 152
  7. 2019Does Object Recognition Work for Everyone?cited by 149
  8. 2023Holistic Analysis of Hallucination in GPT-4V(ision): Bias and Interference Challengescited by 131
  9. 2020Image Representations Learned With Unsupervised Pre-Training Contain Human-like Biasescited by 107
  10. 2021Sustainable Modular Debiasing of Language Modelscited by 98
  11. 2019Predictive Inequity in Object Detectioncited by 77
  12. 2021Worst of Both Worlds: Biases Compound in Pre-trained Vision-and-Language Modelscited by 52
  13. 2020Mitigating Gender Bias in Captioning Systemscited by 50
  14. 2019Exposing and Correcting the Gender Bias in Image Captioning Datasets and Modelscited by 36
  15. 2023'Person' == Light-skinned, Western Man, and Sexualization of Women of Color: Stereotypes in Stable Diffusioncited by 33
  16. 2023Parameter-efficient Modularised Bias Mitigation via AdapterFusioncited by 25
  17. 2023A Friendly Face: Do Text-to-Image Systems Rely on Stereotypes when the Input is Under-Specified?cited by 25
  18. 2018Bleaching Text: Abstract Features for Cross-lingual Gender Predictioncited by 6

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2024: 4; 2025: 5; 2026: 2.

11 papers.

  1. 2024VLBiasBench: A Comprehensive Benchmark for Evaluating Bias in Large Vision-Language Modelcited by 26
  2. 2024Debiasing Large Vision-Language Models by Ablating Protected Attribute Representationscited by 5
  3. 2025Bias in Gender Bias Benchmarks: How Spurious Features Distort Evaluationcited by 5
  4. 2025DSO: Direct Steering Optimization for Bias Mitigationcited by 3
  5. 2025Bias Beyond Demographics: Probing Decision Boundaries in Black-Box LVLMs via Counterfactual VQAcited by 2
  6. 2024debiaSAE: Benchmarking and Mitigating Vision-Language Model Biascited by 1
  7. 2025A Stereotype Content Analysis on Color-related Social Bias in Large Vision Language Modelscited by 1
  8. 2025Surgeons Are Indian Males and Speech Therapists Are White Females: Auditing Biases in Vision-Language Models for Healthcare Professionalscited by 1
  9. 2024Debias your Large Multi-Modal Model at Test-Time via Non-Contrastive Visual Attribute Steeringcited by 0
  10. 2026Guardrail-Agnostic Societal Bias Evaluation in Large Vision-Language Modelscited by 0
  11. 2026LLM-Driven Autonomous Vehicles Inherit Human Driver Biases in Pedestrian Yielding: Results and Implications From A New Benchmarkcited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2024Uncovering Bias in Large Vision-Language Models at Scale with Counterfactuals
  2. 2025Data Matters Most: Auditing Social Bias in Contrastive Vision Language Models
  3. 2024Images Speak Louder than Words: Understanding and Mitigating Bias in Vision-Language Model from a Causal Mediation Perspective
  4. 2023MultiModal Bias: Introducing a Framework for Stereotypical Bias Assessment beyond Gender and Race in Vision Language Models
  5. 2025VIGNETTE: Socially Grounded Bias Evaluation for Vision-Language Models
  6. 2025Bias in the Picture: Benchmarking VLMs with Social-Cue News Images and LLM-as-Judge Assessment
  7. 2025Ask Me Again Differently: GRAS for Measuring Bias in Vision Language Models on Gender, Race, Age, and Skin Tone
  8. 2024BiasDora: Exploring Hidden Biased Associations in Vision-Language Models