Title: Efficient Weighted Sampling via Score-based Generative Models

Original: arXiv:2502.04646v2 Announce Type: replace-cross Abstract: Weighted sampling -- sampling from a probability density function (PDF) proportional to the product of a base PDF and a weight function -- is a fundamental technique with wide-ranging applications in variance reduction, biased sampling, data augmentation, and more. Leveraging the increasing availability of pretrained score-based generative models (SGMs), we propose a training-free weighted sampling framework that approximates the backward diffusion process of the target distribution by augmenting the pretrained base score function with an auxiliary guidance term, in a principled and computationally efficient manner. Our approach builds on two key components: a lightweight approximation of the guidance that avoids costly higher-order derivatives of both the score and weight functions, and an uncertainty-aware scheduler that dynamically adjusts the guidance strength based on a temporal analysis of approximation error. Together, these components enable accurate and stable sampling without relying on particle-based resampling or Hessian evaluations commonly required by existing methods. We validate the effectiveness of our method from synthetic to large-scale settings such as Stable Diffusion XL, where our framework achieves $1.2\times$ to $4.7\times$ speedups while consistently matching or outperforming state-of-the-art baselines in task performance. These results position our method as a scalable and inference-efficient solution for task-adaptive, time-sensitive sampling in generative applications.

Rewrite: Efficient Weighted Sampling via Score-based Generative Models

arXiv:2502.04646v2 Announce Type: replace-cross Abstract: A foundational technique employed across diverse fields such as data augmentation, biased sampling, and variance reduction is weighted sampling, which involves drawing samples from a probability density function (PDF) that is proportional to the product of a base PDF and a specific weight function. Taking advantage of the growing prevalence of pretrained score-based generative models (SGMs), we introduce a novel, training-free framework for weighted sampling. This method efficiently approximates the target distribution’s backward diffusion process by integrating an auxiliary guidance term into the pretrained base score function. The proposed approach relies on two primary innovations: first, a streamlined guidance approximation that sidesteps the need for expensive higher-order derivatives of the weight and score functions; and second, an uncertainty-aware scheduler that modulates guidance intensity dynamically through a temporal assessment of approximation errors. By combining these elements, our method ensures stable and precise sampling, eliminating the necessity for Hessian evaluations or particle-based resampling typically demanded by prior techniques. We demonstrate the efficacy of our approach across various scenarios, ranging from synthetic datasets to large-scale models like Stable Diffusion XL. In these tests, our framework delivers speed improvements ranging from $1.2\times$ to $4.7\times$, while maintaining or surpassing the performance of current state-of-the-art baselines. These findings highlight our method as a robust, scalable, and inference-efficient option for time-critical, task-adaptive sampling within generative applications.

Source: arXiv Generated at: 2026-06-02 00:00:00 UTC