发表机构
Imperial College London（帝国理工学院）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
提出排序感知投机采样（RASS），利用草稿候选的排序信息优化扩散草稿树的验证，在多个生成任务上较 D-GRS 提升约 20% 效率。
AI 中文摘要
投机采样通过并行验证廉价的草稿状态来加速扩散生成，同时保持目标分布不变。最近的基于树的方法比单链草稿更有效地分配并行计算预算，如扩散贪心拒绝采样（D-GRS）所示。D-GRS 在每个节点生成 $K$ 个条件独立的候选，并按生成顺序依次测试它们。然而，采样出的候选无需额外的目标模型评估即可提供有信息量的排序。为了利用这一点，我们引入了排序感知投机采样（RASS），一种基于排序感知列表耦合的投机草稿树验证规则。RASS 沿提议-目标均值位移对草稿候选进行排序，并以最小化所选提议与目标分布之间的总变差为目标，用优化后的权重采样一个排序。最后，所选候选与目标进行最大耦合，残差校正确保对任何排序权重选择都能精确采样。我们在高斯混合目标、FFHQ 上的无条件像素空间生成、CIFAR-10 上的条件生成以及使用 COCO2014 提示的 Stable Diffusion 3.5 潜在扩散上评估了 RASS。以标准采样与投机采样的目标模型评估次数之比衡量，RASS 在评估设置中均优于 D-GRS，在匹配计算预算下在 CIFAR-10 上增益达到约 20%。
英文摘要
Speculative sampling accelerates diffusion generation by verifying inexpensive draft states in parallel while preserving the target law. Recent tree-based methods allocate the parallel compute budget more effectively than single-chain drafts, as demonstrated by Diffusion Greedy Rejection Sampling (D-GRS). D-GRS generates $K$ conditionally independent candidates per node, and sequentially tests them in their generation order. Yet the sampled candidates admit an informative ranking without additional target-model evaluations. To exploit this, we introduce Rank-Aware Speculative Sampling (RASS), a verification rule for speculative draft trees based on rank-aware list coupling. RASS orders draft candidates along the proposal-target mean displacement and samples a rank with weights optimized to minimize total variation between the selected-proposal and target laws. Finally, the selected candidate is maximally coupled with the target, with residual correction ensuring exact sampling for any choice of rank weights. We evaluate RASS on a Gaussian-mixture target, unconditional pixel-space generation on FFHQ, conditional generation on CIFAR-10, and latent diffusion with Stable Diffusion 3.5 using COCO2014 prompts. Measured by the ratio of standard to speculative sampling's target-model evaluation counts, RASS improves on D-GRS across the evaluated settings, with gains reaching approximately 20% on CIFAR-10 at matched compute budgets.
