面相学再审视：视觉语言模型如何通过图像变换编码敏感属性
原标题：Physiognomy Revisited: How Vision-Language Models Encode Sensitive Attributes Through Image Transformation
AI 摘要
该论文研究视觉语言模型（VLM）在图像变换任务中如何编码敏感身份属性。作者发现，当要求模型将人脸图像转换为特定种族（如西班牙裔、黑人、白人、亚裔）或性取向（如同性恋、异性恋）时，模型经常照做，且是否照做具有非确定性、群体敏感性，生成图像在统计上可区分并常强化刻板印象。这些表征偏差还会蔓延到文本推理与人物描述任务，并扩展到犯罪性等敏感属性，揭示了多模态AI系统间接进行画像的风险。
正文节选
Physiognomy Revisited: How Vision–Language Models Encode Sensitive Attributes Through Image Transformation Abstract Emergent generative AI technologies, while remarkably beneficial across several tasks and domains, can suffer from unforeseen vulnerabilities and difficult-to-anticipate failure modes. Such failure modes can disparately affect vulnerable communities. This paper investigates two well-known vision language models (VLMs) and inspects how these models encode protected identities while