mlsanityData to 7 Oct 2026

OASIS Uncovers: High-Quality T2I Models, Same Old Stereotypes

arXiv 2501.00962 ↗First posted 1 Jan 2025cs.CV cs.CY cs.LGcited by 8

Not on the map. Why?

Abstract (arXiv)

Images generated by text-to-image (T2I) models often exhibit visual biases and stereotypes of concepts such as culture and profession. Existing quantitative measures of stereotypes are based on statistical parity that does not align with the sociological definition of stereotypes and, therefore, incorrectly categorizes biases as stereotypes. Instead of oversimplifying stereotypes as biases, we propose a quantitative measure of stereotypes that aligns with its sociological definition. We then propose OASIS to measure the stereotypes in a generated dataset and understand their origins within the T2I model. OASIS includes two scores to measure stereotypes from a generated image dataset: (M1) Stereotype Score to measure the distributional violation of stereotypical attributes, and (M2) WALS to measure spectral variance in the images along a stereotypical attribute. OASIS also includes two methods to understand the origins of stereotypes in T2I models: (U1) StOP to discover attributes that the T2I model internally associates with a given concept, and (U2) SPI to quantify the emergence of stereotypical attributes in the latent space of the T2I model during image generation. Despite the considerable progress in image fidelity, using OASIS, we conclude that newer T2I models such as FLUX.1 and SDv3 contain strong stereotypical predispositions about concepts and still generate images with widespread stereotypical attributes. Additionally, the quantity of stereotypes worsens for nationalities with lower Internet footprints.

Key facts. Cited by 8 papers in the map (data through 7 Oct 2026). Builds on 37 of its 52 references that are in the map. First posted on arXiv 1 Jan 2025.

Builds on

37 of 52 references.

References · 37

  1. 2021Learning Transferable Visual Models From Natural Language Supervisioncited by 33,456
  2. 2021High-Resolution Image Synthesis with Latent Diffusion Modelscited by 18,694
  3. 2024The Llama 3 Herd of Modelscited by 14,192
  4. 2022LAION-5B: An open large-scale dataset for training next generation image-text modelscited by 4,131
  5. 2024Scaling Rectified Flow Transformers for High-Resolution Image Synthesiscited by 4,094
  6. 2014Deep Learning Face Attributes in the Wildcited by 3,679
  7. 2016Man is to Computer Programmer as Woman is to Homemaker? Debiasing Word Embeddingscited by 1,304
  8. 2022Mind the Gap: Understanding the Modality Gap in Multi-modal Contrastive Representation Learningcited by 553
  9. 2022Safe Latent Diffusion: Mitigating Inappropriate Degeneration in Diffusion Modelscited by 501
  10. 2021Multimodal datasets: misogyny, pornography, and malignant stereotypescited by 326
  11. 2022Easily Accessible Text-to-Image Generation Amplifies Demographic Stereotypes at Large Scalecited by 262
  12. 2021Annotators with Attitudes: How Annotator Beliefs And Identities Bias Toxic Language Detectioncited by 258
25 more
  1. 2024Applying Guidance in a Limited Interval Improves Sample and Distribution Quality in Diffusion Modelscited by 240
  2. 2023Stable Bias: Analyzing Societal Representations in Diffusion Modelscited by 225
  3. 2019Are We Modeling the Task or the Annotator? An Investigation of Annotator Bias in Natural Language Understanding Datasetscited by 173
  4. 2021ZeroCap: Zero-Shot Image-to-Text Generation for Visual-Semantic Arithmeticcited by 150
  5. 2023Fair Diffusion: Instructing Text-to-Image Generation Models on Fairnesscited by 149
  6. 2023Debiasing Vision-Language Models via Biased Promptscited by 147
  7. 2021Evaluating CLIP: Towards Characterization of Broader Capabilities and Downstream Implicationscited by 128
  8. 2023Social Biases through the Text-to-Image Generation Lenscited by 126
  9. 2022How well can Text-to-Image Generative Models understand Ethical Natural Language Interventions?cited by 92
  10. 2023The Bias Amplification Paradox in Text-to-Image Generationcited by 69
  11. 2023SeeGULL: A Stereotype Benchmark with Broad Geo-Cultural Coverage Leveraging Generative Modelscited by 66
  12. 2023ITI-GEN: Inclusive Text-to-Image Generationcited by 64
  13. 2024OpenBias: Open-set Bias Detection in Text-to-Image Generative Modelscited by 61
  14. 2023Into the LAIONs Den: Investigating Hate in Multimodal Datasetscited by 54
  15. 2024Balancing Act: Distribution-Guided Debiasing in Diffusion Modelscited by 50
  16. 2024ViSAGe: A Global-Scale Analysis of Visual Stereotypes in Text-to-Image Generationcited by 27
  17. 2023Stereotypes and Smut: The (Mis)representation of Non-cisgender Identities by Text-to-Image Modelscited by 26
  18. 2024FairerCLIP: Debiasing CLIP's Zero-Shot Predictions using Functions in RKHSscited by 25
  19. 2021Measuring Model Biases in the Absence of Ground Truthcited by 24
  20. 2024The Dark Side of Dataset Scaling: Evaluating Racial Classification in Multimodal Modelscited by 23
  21. 2024Utility-Fairness Trade-Offs and How to Find Themcited by 15
  22. 2021On Characterizing the Trade-off in Invariant Representation Learningcited by 12
  23. 2024Fairness and Bias Mitigation in Computer Vision: A Surveycited by 12
  24. 2024Controllable Prompt Tuning For Balancing Group Distributional Robustnesscited by 10
  25. 2024A Taxonomy of the Biases of the Images created by Generative Artificial Intelligencecited by 4

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2025: 5; 2026: 3.

8 papers.

  1. 2025A Large Scale Analysis of Gender Biases in Text-to-Image Generative Modelscited by 12
  2. 2025CoInD: Enabling Logical Compositions in Diffusion Modelscited by 8
  3. 2025Person-Centric Annotations of LAION-400M: Auditing Bias and Its Transfer to Modelscited by 4
  4. 2025Surfacing Subtle Stereotypes: A Multilingual, Debate-Oriented Evaluation of Modern LLMscited by 2
  5. 2026Gender Artifacts from Art History to Text-to-Image Generationcited by 1
  6. 2025Text-to-Image Models and Their Representation of People from Different Nationalities Engaging in Activitiescited by 0
  7. 2026Beyond wheelchairs and blindfolds: Investigating disability stereotypes in T2I models with INCLUDE-BENCHcited by 0
  8. 2026ContextBias: Controlled Evaluation of Bias Persistence Under Context Shift in Text-to-Image Modelscited by 0