自动挖掘文本生成图像模型中的偏见提示，揭示其隐性倾向。
MineTheGap: Automatic Mining of Biases in Text-to-Image Models
- 用遗传算法优化提示词，系统发现易引发偏见的输入组合。
- 通过对比图像与LLM生成文本分布，量化偏见严重程度。
- 可帮助开发者检测并缓解模型在职业、种族等场景的刻板印象。
文本到图像（TTI）模型根据文本提示生成图像，但提示常存在语义模糊，导致模型在解释时产生偏见，可能造成社会影响（如特定职业仅关联某种族）或用户生成结果重复。本文提出MineTheGap方法，自动挖掘能触发模型偏见的提示。该方法不局限于单个提示的偏见检测，而是采用遗传算法迭代优化提示池，寻找暴露偏见的模式。其核心是新型偏见评分机制，基于生成图像分布与相应LLM生成文本变体分布的差异进行排序，并在已知偏见数据集上验证有效性。代码与示例已公开于项目主页。
原文摘要 · Abstract (English)
Text-to-Image (TTI) models generate images based on text prompts, which often leave certain aspects of the desired image ambiguous. When faced with these ambiguities, TTI models have been shown to exhibit biases in their interpretations. These biases can have societal impacts, e.g., when showing only a certain race for a stated occupation. They can also affect user experience when creating redundancy within a set of generated images instead of spanning diverse possibilities. Here, we introduce MineTheGap - a method for automatically mining prompts that cause a TTI model to generate biased outputs. Our method goes beyond merely detecting bias for a given prompt. Rather, it leverages a genetic algorithm to iteratively refine a pool of prompts, seeking for those that expose biases. This optimization process is driven by a novel bias score, which ranks biases according to their severity, as we validate on a dataset with known biases. For a given prompt, this score is obtained by comparing the distribution of generated images to the distribution of LLM-generated texts that constitute variations on the prompt. Code and examples are available on the project's webpage.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。