DefaultShift：审计加速文本到图像模型中的语义默认偏移
DefaultShift: Auditing Semantic Default Shift in Accelerated Text-to-Image Models
浏览论文内容
中文总结 AI 辅助
本研究提出DefaultShift审计方法检测文本到图像模型加速时的语义默认偏移，还提出DefaultShift-Select校准方法降低偏移，提升模型语义保护能力。
中文摘要 AI 辅助
少步文本到图像模型正逐渐取代速度较慢的生成器，然而即使单个输出看起来合理且符合要求，加速过程仍可能悄悄改变未指定属性的分布，我们将这些分布称为语义默认，将其在模型替换下的变化称为语义默认偏移。现有的质量、偏好和多样性评估均未测试替换模型是否保留了参考模型的语义默认。我们提出DefaultShift，这是一种配对审计方法，它使用封闭语义词汇标记重复样本，测量概率质量的移动，并将可解释的排序与验证性交叉拟合推理分离。在14组参考模型与替换模型的配对中，调整后的颜色差异范围为0.054至0.303，且具有特定于方案的方向。对1000张图像的人工审计重现了该排序。我们进一步提出DefaultShift-Select，这是一种离线校准方法，它在不造成实质性质量损失的情况下，将Turbo、DMD2和FLUX模型的人工测量偏移降低了10.3%至35.1%。在平衡评估下，所选数据比未校准的替换数据多恢复了4.3个准确率点和7.5个最差组点。DefaultShift使加速过程中的语义保护变得可测量且可操作。
英文摘要
Few-step text-to-image models increasingly replace slower generators, yet acceleration can silently change distributions over unspecified attributes even when individual outputs remain plausible and aligned. We call these distributions semantic defaults and their change under replacement semantic default shift. Existing quality, preference, and diversity evaluations do not test whether a replacement preserves its reference model's semantic defaults. We introduce DefaultShift, a paired audit that labels repeated samples with closed semantic vocabularies, measures probability-mass movement, and separates interpretable ranking from confirmatory cross-fit inference. Across 14 reference and replacement pairs, adjusted color discrepancies range from 0.054 to 0.303 with recipe-specific directions. A 1,000-image human audit reproduces the ordering. We further introduce DefaultShift-Select, an offline calibration method that reduces human-measured shift by 10.3 percent to 35.1 percent across Turbo, DMD2, and FLUX without material quality loss. Under balanced evaluation, selected data recover 4.3 accuracy points and 7.5 worst-group points over uncalibrated replacement data. DefaultShift makes semantic preservation under acceleration measurable and actionable.
