用一致性模型+重要性采样，6-25次评估即可无偏采样。
Efficient and Unbiased Sampling of Boltzmann Distributions via Consistency Models
- 将一致性模型与重要性采样结合，解决采样偏差问题。
- 仅需6-25次函数评估（NFE），有效样本量媲美100次的DDPM。
- 适合需要高效、高精度采样的物理模拟与生成建模场景。
扩散模型在推进玻尔兹曼生成器方面展现出巨大潜力，但仍有两大挑战：(1) 模型不完美导致样本固有误差；(2) 高质量采样需数百次函数评估（NFE）。现有方法如重要性采样和蒸馏虽分别缓解问题，但常不兼容，因多数蒸馏模型缺乏重要性采样所需密度信息。本文提出一种新采样方法，将一致性模型（CMs）与重要性采样有效结合。我们在合成能量函数及等变多体粒子系统上进行评估。结果表明，该方法仅需6-25 NFE即可生成无偏样本，且有效样本量（ESS）可比肩需约100 NFE的去噪扩散概率模型（DDPM）。
原文摘要 · Abstract (English)
Diffusion models have shown promising potential for advancing Boltzmann Generators. However, two critical challenges persist: (1) inherent errors in samples due to model imperfections, and (2) the requirement of hundreds of functional evaluations (NFEs) to achieve high-quality samples. While existing solutions like importance sampling and distillation address these issues separately, they are often incompatible, as most distillation models lack the necessary density information for importance sampling. This paper introduces a novel sampling method that effectively combines Consistency Models (CMs) with importance sampling. We evaluate our approach on both synthetic energy functions and equivariant n-body particle systems. Our method produces unbiased samples using only 6-25 NFEs while achieving a comparable Effective Sample Size (ESS) to Denoising Diffusion Probabilistic Models (DDPMs) that require approximately 100 NFEs.
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。