提出无训练扩散生成新方法，提升条件生成精度与效率。
Multilevel and Sequential Monte Carlo for Training-Free Diffusion Guidance
- 用序列蒙特卡洛构建无偏后验估计，更好捕捉多模态特性。
- 在CIFAR-10上达95.6%准确率，成功率成本降低3倍。
- 适合追求高精度、低开销的生成模型应用者。
我们解决训练无关条件下扩散模型中精确生成的难题。现有方法通常依赖点估计近似后验得分，常导致偏差，无法捕捉扩散反向过程的多模态本质。本文提出一种序列蒙特卡洛（SMC）框架，通过蒙特卡洛近似对完整去噪分布积分，构建 $p_θ(y|x_t)$ 的无偏估计器。为保证计算可行性，引入基于多级蒙特卡洛（MLMC）的方差缩减策略。该方法在CIFAR-10类别条件生成任务上达到新的最优性能，准确率达95.6%，单位成功成本比基线低3倍；在ImageNet上，成本优势达1.5倍。
原文摘要 · Abstract (English)
We address the problem of accurate, training-free guidance for conditional generation in trained diffusion models. Existing methods typically rely on point-estimates to approximate the posterior score, often resulting in biased approximations that fail to capture multimodality inherent to the reverse process of diffusion models. We propose a sequential Monte Carlo (SMC) framework that constructs an unbiased estimator of $p_θ(y|x_t)$ by integrating over the full denoising distribution via Monte Carlo approximation. To ensure computational tractability, we incorporate variance-reduction schemes based on Multi-Level Monte Carlo (MLMC). Our approach achieves new state-of-the-art results for training-free guidance on CIFAR-10 class-conditional generation, achieving $95.6\%$ accuracy with $3\times$ lower cost-per-success than baselines. On ImageNet, our algorithm achieves $1.5\times$ cost-per-success advantage over existing methods.
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。