发表机构
Peking University（北京大学）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
该研究通过像素级干预审计4种图像评分器，发现评分器的主导效应是保真度偏好，合成审计会误判偏见，仅真实数据的图像内因果隔离可区分真实人口统计学偏见与保真度偏好。
AI 中文摘要
文本到图像系统使用学习得到的美学评分器来过滤训练数据并引导生成，但这些评分是否将人口统计学属性编码为客观质量尚不清楚。我们在合成图像和真实图像上，通过对肤色和体型进行像素级干预，审计了四个评分器：LAION-Aesthetics、PickScore、ImageReward、HPSv2。我们的关键发现是，沿肤色亮度方向，主导效应是保真度偏好：未修改的图像得分最高，且任一方向的扰动都会受到惩罚（呈倒U型）。安慰剂组显示，这种惩罚并非肤色操作符的人为产物，因为对非皮肤区域应用相同的CIELAB L*偏移会产生相似的惩罚幅度。不过，该惩罚依赖于操作符，且仅对LAION-Aes而言对所有操作符成立。重要的是，仅对合成图像进行审计会产生误导：LAION-Aes在合成人脸中表现出对更深肤色的强烈偏好，但在1470张真实人脸中，该偏好发生逆转且幅度大幅减小，放大效应不再显著。在各评分器中，合成结果无法迁移——LAION-Aes和HPSv2的结果发生逆转，PickScore的结果则减弱。我们贡献了一个具有人工制品控制和合成/真实交叉验证的可复现基准，以及一个用于像素级因果隔离的可审计性标准（适用于肤色，但因变形不适用于体型）。按人群分层的分析显示，经FDR校正后，除HPSv2外，保真度惩罚的不对称性在各组间不具稳健性。我们的发现表明，单纯的合成审计会错误判断偏见的方向和幅度，只有在真实数据上进行图像内因果隔离，才能区分真实的人口统计学偏见与保真度偏好。
英文摘要
Text-to-image systems use learned aesthetic scorers to filter training data and guide generation, but whether these scores encode demographic attributes as objective quality is unclear. We audit four scorers (LAION-Aesthetics, PickScore, ImageReward, HPSv2) using pixel-level interventions on skin tone and body type in synthetic and real images. Our key finding is that along skin-lightness, the dominant effect is fidelity preference: unaltered images score highest, and perturbations in either direction are penalized (inverted-U). Placebo arms show this penalty is not an artifact of the skin operator, as applying the same CIELAB L* shift to non-skin regions yields similar penalty magnitudes. However, the penalty is operator-dependent and holds for all operators only for LAION-Aes. Critically, audits on synthetic images alone are misleading: LAION-Aes shows strong preference for darker skin on synthetic faces, but on 1470 real faces the preference reverses and becomes much smaller, and amplification becomes non-significant. Across scorers, synthetic results do not transfer -- reversing for LAION-Aes and HPSv2, attenuating for PickScore. We contribute a reproducible benchmark with artifact control and synthetic/real cross-validation, and an auditability criterion for pixel-level causal isolation (valid for skin tone, not for body type due to deformation). Population-stratified analysis shows fidelity-penalty asymmetry is not robust across groups after FDR correction except for HPSv2. Our findings show naive synthetic audits misjudge bias direction and magnitude, and only within-image causal isolation on real data can distinguish true demographic bias from fidelity preference.
