发表机构
The Scripps Research Institute（斯克里普斯研究所）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
提出无需训练的GeoSPRINT框架，通过检测扩散轨迹几何冗余构建非均匀采样调度，在多个数据集上较DDIM等方法提升采样效率与FID，无需重新训练。
AI 中文摘要
扩散模型虽能实现高样本质量，但推理时仍存在高开销问题，原因是采样需要多次顺序的神经函数评估（NFE）。现有加速方法要么采用固定的跳步调度，要么基于局部数值误差调整步长，要么需要额外训练。我们提出GeoSPRINT（面向轨迹推理的几何步长剪枝），这是一种无需训练的框架，可从去噪轨迹的几何结构构建非均匀采样调度。GeoSPRINT通过潜在空间中的超平面测试检测几何冗余步长，该测试通过QR分解高效实现，并将所得冗余分布转换为采样调度，为轨迹的高曲率区域分配更多步长。此外，我们提出轨迹投影分数α_traj，这是一种残差方差度量，用于量化轨迹的平直度，并作为整流流质量的无模型诊断指标。在CIFAR-10（32×32）、LSUN Church（256×256）和Stable Diffusion v1.5（512×512潜在空间）上，GeoSPRINT在匹配的NFE预算下始终优于均匀DDIM（去噪扩散隐式模型）调度。在CIFAR-10上，GeoSPRINT在49-89个NFE下较DDIM将FID（Fréchet Inception距离）提升0.7-1.1，且尽管使用一阶DDIM求解器，在NFE≥30时仍优于DPM-Solver++。在LSUN Church上，其在52步时将FID从1.48降至1.26；在Stable Diffusion v1.5上，其较DDIM实现高达1.93的FID提升。这些结果表明，轨迹几何结构为分配推理步长提供了有用的全局信号，且调度质量可在无需重新训练的情况下大幅提升扩散采样效率。
英文摘要
Diffusion models achieve high sample quality but remain expensive at inference time because sampling requires many sequential neural function evaluations (NFEs). Existing acceleration methods either use fixed step-skipping schedules, adapt step sizes based on local numerical error, or require additional training. We introduce GeoSPRINT (Geometric Step Pruning for Inference in Trajectories), a training-free framework for constructing non-uniform sampling schedules from the geometry of denoising trajectories. GeoSPRINT detects geometrically redundant steps using a hyperplanarity test in latent space, implemented efficiently via QR factorization, and converts the resulting redundancy profile into a sampling schedule that allocates more steps to high-curvature regions of the trajectory. In addition, we introduce the trajectory projection score $α_{\mathrm{traj}}$, a residual-variance metric that quantifies trajectory straightness and serves as a model-free diagnostic for rectified flow quality. Across CIFAR-10 ($32{\times}32$), LSUN Church ($256{\times}256$), and Stable Diffusion v1.5 ($512{\times}512$ latent), GeoSPRINT consistently improves over uniform DDIM (Denoising Diffusion Implicit Models) schedules at matched NFE budgets. On CIFAR-10, GeoSPRINT improves FID (Fréchet Inception Distance) by 0.7-1.1 over DDIM across 49-89 NFEs and surpasses DPM-Solver++ at NFE${\geq}30$ despite using a first-order DDIM solver. On LSUN Church, it reduces FID from 1.48 to 1.26 at 52 steps, and on Stable Diffusion v1.5 it achieves up to 1.93 FID improvement over DDIM. These results show that trajectory geometry provides a useful global signal for allocating inference steps and that schedule quality can substantially improve diffusion sampling efficiency without retraining.
