用于文本到图像编辑的 Delta 整流流采样
Delta Rectified Flow Sampling for Text-to-Image Editing
DRFS 提出一种基于蒸馏的无反转编辑方法，通过显式建模源与目标速度场差异并引入时间相关偏移项，缓解过平滑伪影，可复用于 rectified flow 模型的文本引导编辑，无需架构修改。
原文依据 We propose Delta Rectified Flow Sampling (DRFS), a novel inversion-free, path-aware editing framework within rectified flow models for text-to-image editing.
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论我们提出 Delta 整流流采样（DRFS），一种新颖的、无反转的、路径感知的编辑框架，用于整流流模型中的文本到图像编辑。DRFS 是一种基于蒸馏的方法，它显式地建模源速度场与目标速度场之间的差异，以缓解先前蒸馏采样方法中普遍存在的过平滑伪影。我们进一步引入一个时间相关的偏移项，将噪声潜变量推向目标轨迹，增强与目标分布的对齐。我们从理论上证明，禁用该偏移项可恢复 Delta 去噪分数（DDS），从而桥接基于分数的扩散优化与基于速度的整流流优化。此外，在整流流动力学下，线性偏移调度将无反转方法 FlowEdit 恢复为严格的特殊情况，从而给出优化与 ODE 编辑的统一视角。我们进行了一项分析以指导偏移项的设计，并且在广泛使用的 PIE 基准上的实验结果表明，DRFS 在无需架构修改的情况下实现了卓越的编辑质量、保真度和可控性。代码可在 https://github.com/Harvard-AI-and-Robotics-Lab/DeltaRectifiedFlowSampling 获取。
英文原文摘要 1259 字符
We propose Delta Rectified Flow Sampling (DRFS), a novel inversion-free, path-aware editing framework within rectified flow models for text-to-image editing. DRFS is a distillation-based method that explicitly models the discrepancy between the source and target velocity fields in order to mitigate over-smoothing artifacts rampant in prior distillation sampling approaches. We further introduce a time-dependent shift term to push noisy latents closer to the target trajectory, enhancing the alignment with the target distribution. We theoretically demonstrate that disabling this shift recovers Delta Denoising Score (DDS), bridging score-based diffusion optimization and velocity-based rectified-flow optimization. Moreover, under rectified-flow dynamics, a linear shift schedule recovers the inversion-free method FlowEdit as a strict special case, yielding a unifying view of optimization and ODE editing. We conduct an analysis to guide the design of our shift term, and experimental results on the widely used PIE Benchmark indicate that DRFS achieves superior editing quality, fidelity, and controllability while requiring no architectural modifications. Code is available at https://github.com/Harvard-AI-and-Robotics-Lab/DeltaRectifiedFlowSampling.