让图像生成模型保留对初始噪声的敏感度，提升可控性
Restoring Initial Noise Sensitivity in Text-to-Image Distillation via Geometric Alignment
- 通过匹配师生模型对噪声扰动的微分响应，保持生成过程的几何结构
- 在多个生成任务中恢复了原始模型的噪声敏感性，同时保持高图像质量
- 适合需要精细控制生成过程的研究者，如基于噪声的编辑与优化
生成式蒸馏通过将多步生成轨迹压缩为少步学生模型，显著加速文本到图像（T2I）生成，同时保持感知质量。然而，现有方法主要关注效率和输出保真度，常忽略原始轨迹的关键特性。本文发现一个重要缺失：对初始噪声的敏感性退化，影响依赖噪声优化的下游控制方法。我们溯源至标准蒸馏目标中的逐点输出对齐，导致输入-输出景观被平坦化，抑制了教师模型的局部几何结构。为此，提出几何感知蒸馏（GAD），一种保持敏感性的框架，通过匹配输入噪声的雅可比-向量乘积，使学生模型复现教师对扰动的微分响应。大量实验表明，GAD在多种T2I范式和噪声驱动控制任务中显著恢复敏感性并提升多样性，同时维持高视觉保真度。代码已开源。
原文摘要 · Abstract (English)
Generative distillation significantly accelerates text-to-image (T2I) generation by compressing multi-step trajectories into few-step student models while preserving perceptual quality. However, existing methods primarily optimize efficiency and output fidelity, often neglecting critical properties of the original trajectory. In this work, we identify a key missing property: sensitivity to initial noise, whose degradation impairs downstream control methods relying on noise-based optimization and manipulation. We trace this issue to standard distillation objectives that enforce pointwise output alignment, inadvertently flattening the input-output landscape and suppressing the teacher's local geometric structure. To address this, we propose Geometry-Aware Distillation (GAD), a sensitivity-preserving framework that aligns the local functional behavior of teacher and student models. Specifically, GAD matches Jacobian-vector products with respect to input noise, enabling the student to reproduce the teacher's differential response to perturbations. Extensive experiments across multiple T2I paradigms and noise-driven control tasks demonstrate that GAD significantly restores sensitivity and improves diversity while maintaining high visual fidelity. Code is available at https://github.com/Hannah1102/GAD.
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。