mlsanityData to 7 Oct 2026

Social bias and fairness in VLMs

123 papers0 this weekMedium confidence in this grouping: mostly one topic, with blurry edgesMedium confidence in this grouping: mostly one topic, with blurry edges · About

Keywords: bias, gender, biases, social, fairness

Key facts. Social bias and fairness in VLMs: 123 papers on arXiv since 2022 (data through 7 Oct 2026). 24 of its 123 mapped papers (20%) cite “DeAR: Debiasing Vision-Language Models with Additive Residuals” (2023, arXiv 2303.10431). 24 of its 123 mapped papers (20%) cite “Examining Gender and Racial Bias in Large Vision-Language Models Using a Novel Dataset of Parallel Images” (2024, arXiv 2402.05779). 4 new papers 17 Sep–7 Oct.

What’s new

17 Sep – 7 Oct 2026

  1. Earlier · 4
  2. Similar Choices, Different Attention: Cross-Modal Associations in Humans and Vision-Language Models also fits: VLM benchmarks and visual reasoning
  3. When Words Speak Louder than Images: Towards Understanding Language Bias in Vision-Language Models
  4. Gender Bias in Vision-Language In-Context Learning
  5. Feed the Panel Dimensions, Not Verdicts: Rubric-Decomposed Fusion of Vision-Language Aesthetic Judges also fits: VLM benchmarks and visual reasoning

Foundations

Most cited by papers here

  1. DeAR: Debiasing Vision-Language Models with Additive Residuals2023cited by 24 here
  2. Examining Gender and Racial Bias in Large Vision-Language Models Using a Novel Dataset of Parallel Images2024cited by 24 here
  3. A Prompt Array Keeps the Bias Away: Debiasing Vision-Language Models with Adversarial Learning2022cited by 23 here
  4. SocialCounterfactuals: Probing and Mitigating Intersectional Social Biases in Vision-Language Models with Counterfactual Examples2023cited by 22 here
  5. Debiasing Vision-Language Models via Biased Prompts2023cited by 21 here

Lineage

The papers this direction builds on, oldest first.

Each paper on the line cites the one before.

  1. 2022A Prompt Array Keeps the Bias Away: Debiasing Vision-Language Models with Adversarial Learningcited by 23 here
  2. 2023DeAR: Debiasing Vision-Language Models with Additive Residualscited by 24 here
  3. 2024Examining Gender and Racial Bias in Large Vision-Language Models Using a Novel Dataset of Parallel Imagescited by 24 here
  4. 2024A Unified Framework and Dataset for Assessing Societal Bias in Vision-Language Modelscited by 8 here
  5. 2024VLBiasBench: A Comprehensive Benchmark for Evaluating Bias in Large Vision-Language Modelcited by 7 here
19 more it builds on, not in this chain
  1. 2022BLIP: Bootstrapping Language-Image Pre-training for Unified Vision-Language Understanding and Generationcited by 11 here
  2. 2022Counterfactually Measuring and Eliminating Social Bias in Vision-Language Pre-training Modelscited by 9 here
  3. 2023Vision-Language Models Performing Zero-Shot Tasks Exhibit Gender-based Disparitiescited by 7 here
  4. 2023Debiasing Vision-Language Models via Biased Promptscited by 21 here
  5. 2023MultiModal Bias: Introducing a Framework for Stereotypical Bias Assessment beyond Gender and Race in Vision Language Modelscited by 12 here
  6. 2023Vision-Language Models for Vision Tasks: A Surveycited by 9 here
  7. 2023InstructBLIP: Towards General-purpose Vision-Language Models with Instruction Tuningcited by 12 here
  8. 2023VisoGender: A dataset for benchmarking gender bias in image-text pronoun resolutioncited by 19 here
  9. 2023Survey of Social Bias in Vision-Language Modelscited by 8 here
  10. 2023Improved Baselines with Visual Instruction Tuningcited by 20 here
  11. 2023SocialCounterfactuals: Probing and Mitigating Intersectional Social Biases in Vision-Language Models with Counterfactual Examplescited by 22 here
  12. 2023InternVL: Scaling up Vision Foundation Models and Aligning for Generic Visual-Linguistic Taskscited by 7 here
  13. 2024FairerCLIP: Debiasing CLIP's Zero-Shot Predictions using Functions in RKHSscited by 8 here
  14. 2024Uncovering Bias in Large Vision-Language Models at Scale with Counterfactualscited by 9 here
  15. 2024GenderBias-VL: Benchmarking Gender Bias in Vision Language Models via Counterfactual Probingcited by 10 here
  16. 2024Images Speak Louder than Words: Understanding and Mitigating Bias in Vision-Language Model from a Causal Mediation Perspectivecited by 7 here
  17. 2024LLaVA-OneVision: Easy Visual Task Transfercited by 7 here
  18. 2024Qwen2-VL: Enhancing Vision-Language Model's Perception of the World at Any Resolutioncited by 11 here
  19. 2024Identifying Implicit Social Biases in Vision-Language Modelscited by 12 here

The 24 most cited of 270.

Browse

30 of 123, ranked 6–35

  1. 6VisoGender: A dataset for benchmarking gender bias in image-text pronoun resolution202319
  2. 7Identifying Implicit Social Biases in Vision-Language Models202412
  3. 8MultiModal Bias: Introducing a Framework for Stereotypical Bias Assessment beyond Gender and Race in Vision Language Models202312
  4. 9GenderBias-VL: Benchmarking Gender Bias in Vision Language Models via Counterfactual Probing202410
  5. 10Uncovering Bias in Large Vision-Language Models at Scale with Counterfactuals20249
  6. 11Counterfactually Measuring and Eliminating Social Bias in Vision-Language Pre-training Models20229
  7. 12A Unified Framework and Dataset for Assessing Societal Bias in Vision-Language Models20248
  8. 13Survey of Social Bias in Vision-Language Models20238
  9. 14VLBiasBench: A Comprehensive Benchmark for Evaluating Bias in Large Vision-Language Model20247
  10. 15Images Speak Louder than Words: Understanding and Mitigating Bias in Vision-Language Model from a Causal Mediation Perspective20247
All 30
  1. 16BendVLM: Test-Time Debiasing of Vision-Language Embeddings20247
  2. 17Vision-Language Models Performing Zero-Shot Tasks Exhibit Gender-based Disparities20237
  3. 18Evaluating Fairness in Large Vision-Language Models Across Diverse Demographic Attributes and Prompts20246
  4. 19A Unified Debiasing Approach for Vision-Language Models across Modalities and Tasks20246
  5. 20FaceXBench: Evaluating Multimodal LLMs on Face Understanding20255
  6. 21VisBias: Measuring Explicit and Implicit Social Biases in Vision Language Models20255
  7. 22Uncovering Bias in Large Vision-Language Models with Counterfactuals20245
  8. 23BiasDora: Exploring Hidden Biased Associations in Vision-Language Models20245
  9. 24SANER: Annotation-free Societal Attribute Neutralizer for Debiasing CLIP20245
  10. 25Balancing the Picture: Debiasing Vision-Language Datasets with Synthetic Contrast Sets20235
  11. 26FaceBench: A Multi-View Multi-Level Facial Attribute VQA Dataset for Benchmarking Face Perception MLLMs20254
  12. 27FaceLLM: A Multimodal Large Language Model for Face Understanding20254
  13. 28FairDeDup: Detecting and Mitigating Vision-Language Fairness Disparities in Semantic Dataset Deduplication20244
  14. 29Interpretable Debiasing of Vision-Language Models for Social Fairness20263
  15. 30Unveiling the "Fairness Seesaw": Discovering and Mitigating Gender and Race Bias in Vision-Language Models20253
  16. 31PRISM: Reducing Spurious Implicit Biases in Vision-Language Models with LLM-Guided Embedding Projection20253
  17. 32Social Perception of Faces in a Vision-Language Model20243
  18. 33Face-MLLM: A Large Face Perception Model20243
  19. 34Intrinsic Bias is Predicted by Pretraining Data and Correlates with Downstream Performance in Vision-Language Encoders20252
  20. 35Ask Me Again Differently: GRAS for Measuring Bias in Vision Language Models on Gender, Race, Age, and Skin Tone20252