发表机构
Massachusetts Institute of Technology（麻省理工学院）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
针对扩散模型特异性感知生成中负参考分布与正分布重叠导致质量退化的问题，提出基于重叠目标的目标设计及方差缩减顺序蒙特卡洛采样方法，有效抑制非期望区域并提升采样稳定性。
AI 中文摘要
推理时引导使得预训练的扩散模型能够在不进行完全重新训练的情况下满足新的约束条件。然而，特异性感知生成是困难的：从负参考分布中排斥样本也可能侵蚀正分布中两者重叠的部分。关键挑战是在最小化扭曲正分布的同时抑制负质量。我们通过将特异性感知引导表述为一个目标设计问题，并从一个基于重叠的目标中推导出目标分布来解决这个问题。所得目标仅在期望参考分布相对于非期望参考分布被充分偏好的区域保留该期望参考分布，从而给出了特异性的似然比解释。为了从相应的时间依赖目标路径中采样，我们开发了一个具有方差最小化局部提议的顺序蒙特卡洛采样器。我们进一步引入了一种实用的固定噪声优化程序，该程序使用期望和非期望分数场的雅可比-向量积。在合成任务、类别对比生成、文本到图像任务以及肽-主要组织相容性复合体（p-MHC）结合物上的实验表明，与负引导基线相比，所提出的方法更有效地抑制了非期望区域，减少了模式偏移，并通过降低SMC权重坍缩提高了采样稳定性。代码可在以下网址获取：此https URL
英文摘要
Inference-time steering enables pretrained diffusion models to satisfy new constraints without full retraining. However, specificity-aware generation is difficult: repelling samples from a negative reference distribution can also erode the positive distribution where the two overlap. The key challenge is to suppress negative mass while minimally distorting the positive distribution. We address this problem by formulating specificity-aware steering as a target-design problem and deriving a target distribution from an overlap-based objective. The resulting target keeps the desired reference distribution only in regions where it is sufficiently preferred over the undesired reference distribution, giving a likelihood-ratio interpretation of specificity. To sample from the corresponding time-dependent target path, we develop a Sequential Monte Carlo sampler with a variance-minimized local proposal. We further introduce a practical fixed-noise optimization procedure with the Jacobian--vector products with the desired and undesired score fields. Experiments on synthetic task, class-contrastive generation, text-to-image tasks and peptide-MHC (p-MHC) binder show that the proposed method suppresses undesired regions more effectively, reduces mode shift, and improves sampling stability by decreasing the SMC weight collapse compared with negative-guidance baselines. Code is available at: https://github.com/WangLuran/Specificity-Aware-Diffusion-Steering
