# Word-Level Explanations for Analyzing Bias in Text-to-Image Models

arXiv: https://arxiv.org/abs/2306.05500 · HTML: https://mlsanity.com/paper/2306.05500/

## Key facts

- Cited by 8 papers in the map (data through 14 Oct 2026).
- Builds on 10 of its 17 references that are in the map.
- First posted on arXiv 3 Jun 2023.

## Builds on (most cited first, 10 of 10)

- [High-Resolution Image Synthesis with Latent Diffusion Models](https://mlsanity.com/paper/2112.10752/) (2021)
- [A Unified Approach to Interpreting Model Predictions](https://mlsanity.com/paper/1705.07874/) (2017)
- ["Why Should I Trust You?": Explaining the Predictions of Any Classifier](https://mlsanity.com/paper/1602.04938/) (2016)
- [Deep Learning Face Attributes in the Wild](https://mlsanity.com/paper/1411.7766/) (2014)
- [Easily Accessible Text-to-Image Generation Amplifies Demographic Stereotypes at Large Scale](https://mlsanity.com/paper/2211.03759/) (2022)
- [Explaining by Removing: A Unified Framework for Model Explanation](https://mlsanity.com/paper/2011.14878/) (2020)
- [You Only Look at One Sequence: Rethinking Transformer in Vision through Object Detection](https://mlsanity.com/paper/2106.00666/) (2021)
- [Adapting Pretrained Vision-Language Foundational Models to Medical Imaging Domains](https://mlsanity.com/paper/2210.04133/) (2022)
- [No Token Left Behind: Explainability-Aided Image Classification and Generation](https://mlsanity.com/paper/2204.04908/) (2022)
- [Gender and Racial Diversity in Commercial Brands' Advertising Images on Social Media](https://mlsanity.com/paper/1908.01352/) (2019)

## Built on by (most cited first, 8 of 8)

- [Survey of Bias In Text-to-Image Generation: Definition, Evaluation, and Mitigation](https://mlsanity.com/paper/2404.01030/) (2024)
- [Stable Diffusion Exposed: Gender Bias from Prompt to Image](https://mlsanity.com/paper/2312.03027/) (2023)
- [Securing the Future of GenAI: Policy and Technology](https://mlsanity.com/paper/2407.12999/) (2024)
- [Multi-Group Proportional Representation in Retrieval](https://mlsanity.com/paper/2407.08571/) (2024)
- [GradBias: Unveiling Word Influence on Bias in Text-to-Image Generative Models](https://mlsanity.com/paper/2408.16700/) (2024)
- [How Robust is Model Editing after Fine-Tuning? An Empirical Study on Text-to-Image Diffusion Models](https://mlsanity.com/paper/2506.18428/) (2025)
- [Visual Stereotypes of Autism Spectrum in Janus-Pro-7B, DALL-E, Stable Diffusion, SDXL, FLUX, and Midjourney](https://mlsanity.com/paper/2407.16292/) (2024)
- [Testing chatbots on the creation of encoders for audio conditioned image generation](https://mlsanity.com/paper/2509.09717/) (2025)

## Abstract (arXiv)

Text-to-image models take a sentence (i.e., prompt) and generate images associated with this input prompt. These models have created award wining-art, videos, and even synthetic datasets. However, text-to-image (T2I) models can generate images that underrepresent minorities based on race and sex. This paper investigates which word in the input prompt is responsible for bias in generated images. We introduce a method for computing scores for each word in the prompt; these scores represent its influence on biases in the model's output. Our method follows the principle of explaining by removing, leveraging masked language models to calculate the influence scores. We perform experiments on Stable Diffusion to demonstrate that our method identifies the replication of societal stereotypes in generated images.