mlsanityData to 7 Oct 2026

Debiasing Vision-Language Models via Biased Prompts

arXiv 2302.00070 ↗First posted 31 Jan 2023cs.LG cs.CVcited by 147

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

Abstract (arXiv)

Machine learning models have been shown to inherit biases from their training datasets. This can be particularly problematic for vision-language foundation models trained on uncurated datasets scraped from the internet. The biases can be amplified and propagated to downstream applications like zero-shot classifiers and text-to-image generative models. In this study, we propose a general approach for debiasing vision-language foundation models by projecting out biased directions in the text embedding. In particular, we show that debiasing only the text embedding with a calibrated projection matrix suffices to yield robust classifiers and fair generative models. The proposed closed-form solution enables easy integration into large-scale pipelines, and empirical results demonstrate that our approach effectively reduces social bias and spurious correlation in both discriminative and generative vision-language models without the need for additional data or training.

Key facts. Cited by 147 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 46 of its 57 references that are in the map. First posted on arXiv 31 Jan 2023.

Builds on

46 of 57 references.

Closest ancestors · 2

Same direction, newest first

  1. 2023Vision-Language Models Performing Zero-Shot Tasks Exhibit Gender-based Disparities26 Jan
  2. 2022A Prompt Array Keeps the Bias Away: Debiasing Vision-Language Models with Adversarial Learning22 Mar

Other references · 44

  1. 2015Deep Residual Learning for Image Recognitioncited by 51,809
  2. 2021Learning Transferable Visual Models From Natural Language Supervisioncited by 33,456
  3. 2020An Image is Worth 16x16 Words: Transformers for Image Recognition at Scalecited by 29,915
  4. 2021High-Resolution Image Synthesis with Latent Diffusion Modelscited by 18,694
  5. 2022Hierarchical Text-Conditional Image Generation with CLIP Latentscited by 6,539
  6. 2022Photorealistic Text-to-Image Diffusion Models with Deep Language Understandingcited by 6,322
  7. 2022LAION-5B: An open large-scale dataset for training next generation image-text modelscited by 4,131
  8. 2014Deep Learning Face Attributes in the Wildcited by 3,679
  9. 2019A Survey on Bias and Fairness in Machine Learningcited by 1,911
  10. 2016Man is to Computer Programmer as Woman is to Homemaker? Debiasing Word Embeddingscited by 1,304
  11. 2019Distributionally Robust Neural Networks for Group Shifts: On the Importance of Regularization for Worst-Case Generalizationcited by 1,232
  12. 2021CLIP-Adapter: Better Vision-Language Models with Feature Adapterscited by 1,021
32 more
  1. 2021Ethical and social risks of harm from Language Modelscited by 957
  2. 2020Language (Technology) is Power: A Critical Survey of "Bias" in NLPcited by 951
  3. 2020StereoSet: Measuring stereotypical bias in pretrained language modelscited by 911
  4. 2021Robust fine-tuning of zero-shot modelscited by 785
  5. 2022Fine-Tuning can Distort Pretrained Features and Underperform Out-of-Distributioncited by 722
  6. 2017Men Also Like Shopping: Reducing Gender Bias Amplification using Corpus-level Constraintscited by 420
  7. 2022Last Layer Re-Training is Sufficient for Robustness to Spurious Correlationscited by 379
  8. 2018Learning Adversarially Fair and Transferable Representationscited by 353
  9. 2021Multimodal datasets: misogyny, pornography, and malignant stereotypescited by 326
  10. 2019Mitigating Gender Bias in Natural Language Processing: Literature Reviewcited by 287
  11. 2019Lipstick on a Pig: Debiasing Methods Cover up Systematic Gender Biases in Word Embeddings But do not Remove Themcited by 268
  12. 2019Towards Fairness in Visual Recognition: Effective Strategies for Bias Mitigationcited by 252
  13. 2017Word Embeddings Quantify 100 Years of Gender and Ethnic Stereotypescited by 251
  14. 2019Gender Bias in Contextualized Word Embeddingscited by 203
  15. 2022Fine-tuning language models to find agreement among humans with diverse preferencescited by 189
  16. 2020Towards Debiasing Sentence Representationscited by 170
  17. 2018FairGAN: Fairness-aware Generative Adversarial Networkscited by 164
  18. 2019Reducing Sentiment Bias in Language Models via Counterfactual Evaluationcited by 148
  19. 2019Black is to Criminal as Caucasian is to Police: Detecting and Removing Multiclass Bias in Word Embeddingscited by 146
  20. 2019Mitigate Bias in Face Recognition using Skewness-Aware Reinforcement Learningcited by 145
  21. 2022On Feature Learning in the Presence of Spurious Correlationscited by 145
  22. 2019Fairness-Aware Ranking in Search & Recommendation Systems with Application to LinkedIn Talent Searchcited by 137
  23. 2021Evaluating CLIP: Towards Characterization of Broader Capabilities and Downstream Implicationscited by 128
  24. 2020Fair Attribute Classification through Latent Space De-biasingcited by 125
  25. 2018Turning a Blind Eye: Explicit Removal of Biases and Variation from Deep Neural Network Embeddingscited by 114
  26. 2019Fair Generative Modeling via Weak Supervisioncited by 114
  27. 2021Fair Mixup: Fairness via Interpolationcited by 109
  28. 2021Are Gender-Neutral Queries Really Gender-Neutral? Mitigating Gender Bias in Image Searchcited by 90
  29. 2019FairFace: Face Attribute Dataset for Balanced Race, Gender, and Agecited by 87
  30. 2019Bias Correction of Learned Generative Models using Likelihood-Free Importance Weightingcited by 70
  31. 2022Contrastive Adapters for Foundation Model Group Robustnesscited by 62
  32. 2021Measuring Fairness in Generative Modelscited by 31

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2023: 24; 2024: 52; 2025: 44; 2026: 27.

147 papers.

Sort:

By area:

  1. 2023Unified Concept Editing in Diffusion Modelscited by 377
  2. 2023What does CLIP know about a red circle? Visual prompt engineering for VLMscited by 215
  3. 2024Aligning Modalities in Vision Large Language Models via Preference Fine-tuningcited by 197
  4. 2023A Categorical Archive of ChatGPT Failurescited by 167
  5. 2023Finetuning Text-to-Image Diffusion Models for Fairnesscited by 90
  6. 2024Large-scale Reinforcement Learning for Diffusion Modelscited by 78
  7. 2023GraphAdapter: Tuning Vision-Language Models With Dual Knowledge Graphcited by 69
  8. 2023Self-Discovering Interpretable Diffusion Latent Directions for Responsible Text-to-Image Generationcited by 69
  9. 2023ITI-GEN: Inclusive Text-to-Image Generationcited by 64
  10. 2024The Neglected Tails in Vision-Language Modelscited by 63
  11. 2024Examining Gender and Racial Bias in Large Vision-Language Models Using a Novel Dataset of Parallel Imagescited by 55
  12. 2023SocialCounterfactuals: Probing and Mitigating Intersectional Social Biases in Vision-Language Models with Counterfactual Examplescited by 53
38 more
  1. 2023VisoGender: A dataset for benchmarking gender bias in image-text pronoun resolutioncited by 52
  2. 2024Balancing Act: Distribution-Guided Debiasing in Diffusion Modelscited by 50
  3. 2023Linear Spaces of Meanings: Compositional Structures in Vision-Language Modelscited by 48
  4. 2023Inspecting the Geographical Representativeness of Images from Text-to-Image Modelscited by 44
  5. 2024VBench++: Comprehensive and Versatile Benchmark Suite for Video Generative Modelscited by 42
  6. 2023Spawrious: A Benchmark for Fine Control of Spurious Correlation Biasescited by 41
  7. 2024Token Merging for Training-Free Semantic Binding in Text-to-Image Synthesiscited by 35
  8. 2023The Hidden Language of Diffusion Modelscited by 34
  9. 2023Fair Text-to-Image Diffusion via Fair Mappingcited by 31
  10. 2023Balancing the Picture: Debiasing Vision-Language Datasets with Synthetic Contrast Setscited by 27
  11. 2023Survey of Social Bias in Vision-Language Modelscited by 26
  12. 2024B-AVIBench: Towards Evaluating the Robustness of Large Vision-Language Model on Black-box Adversarial Visual-Instructionscited by 26
  13. 2024FairerCLIP: Debiasing CLIP's Zero-Shot Predictions using Functions in RKHSscited by 25
  14. 2023Zero-Shot Robustification of Zero-Shot Modelscited by 24
  15. 2024Would Deep Generative Models Amplify Bias in Future Models?cited by 24
  16. 2024Improving Geo-diversity of Generated Images with Contextualized Vendi Score Guidancecited by 24
  17. 2024Rethinking Training for De-biasing Text-to-Image Generation: Unlocking the Potential of Stable Diffusioncited by 22
  18. 2024Image Generation Diversity Issues and How to Tame Themcited by 22
  19. 2025Prompt-based Adaptation in Large-scale Vision Models: A Surveycited by 21
  20. 2023Mitigating Test-Time Bias for Fair Image Retrievalcited by 20
  21. 2024SANER: Annotation-free Societal Attribute Neutralizer for Debiasing CLIPcited by 19
  22. 2024A Unified Debiasing Approach for Vision-Language Models across Modalities and Taskscited by 18
  23. 2024BendVLM: Test-Time Debiasing of Vision-Language Embeddingscited by 17
  24. 2024Joint Vision-Language Social Bias Removal for CLIPcited by 17
  25. 2023PerceptionCLIP: Visual Classification by Inferring and Conditioning on Contextscited by 16
  26. 2024Multilingual Text-to-Image Generation Magnifies Gender Stereotypes and Prompt Engineering May Not Help Youcited by 15
  27. 2025Personalize Your LLM: Fake it then Align itcited by 15
  28. 2025Debiasing Multimodal Large Language Models via Noise-Aware Preference Optimizationcited by 14
  29. 2023A Systematic Survey of Prompt Engineering on Vision-Language Foundation Modelscited by 12
  30. 2023Evaluating the Fairness of Discriminative Foundation Models in Computer Visioncited by 12
  31. 2024Fairness and Bias Mitigation in Computer Vision: A Surveycited by 12
  32. 2024Defending Multimodal Backdoored Models by Repulsive Visual Prompt Tuningcited by 12
  33. 2025Erasing More Than Intended? How Concept Erasure Degrades the Generation of Non-Target Conceptscited by 12
  34. 2024FairDeDup: Detecting and Mitigating Vision-Language Fairness Disparities in Semantic Dataset Deduplicationcited by 11
  35. 2024MIST: Mitigating Intersectional Bias with Disentangled Cross-Attention Editing in Text-to-Image Diffusion Modelscited by 10
  36. 2024Adaptive Concept Bottleneck for Foundation Models Under Distribution Shiftscited by 9
  37. 2025Project-Probe-Aggregate: Efficient Fine-Tuning for Group Robustnesscited by 9
  38. 2025Fair Generation without Unfair Distortions: Debiasing Text-to-Image Generation with Entanglement-Free Attentioncited by 9

50 most cited of 147 All 147 (JSON)

Similar, not cited

Close in wording, no citation either way

  1. 2024debiaSAE: Benchmarking and Mitigating Vision-Language Model Bias
  2. 2023DeAR: Debiasing Vision-Language Models with Additive Residuals
  3. 2025Freeze and Reveal: Exposing Modality Bias in Vision-Language Models
  4. 2023GELDA: A generative language annotation framework to reveal visual biases in datasets
  5. 2026Bias Detection and Rotation-Robustness Mitigation in Vision-Language Models and Generative Image Models
  6. 2025Biases Propagate in Encoder-based Vision-Language Models: A Systematic Analysis From Intrinsic Measures to Zero-shot Retrieval Outcomes
  7. 2023MultiModal Bias: Introducing a Framework for Stereotypical Bias Assessment beyond Gender and Race in Vision Language Models
  8. 2026Language Bias in LVLMs: From In-Depth Analysis to Simple and Effective Mitigation