arXiv：cs.LG（机器学习，全量分类）· Luran Wang, Linrui Ma, Hannes St\"ark, Regina Barzilay·· 14 小时前AI 评分33
基于方差缩减序列蒙特卡洛的特定性感知扩散引导
Specificity-Aware Diffusion Steering via Variance-Reduced Sequential Monte Carlo
AI 导读
研究者提出一种特定性感知扩散引导方法，将引导建模为目标分布设计问题，基于重叠度目标推导出仅在期望参考分布占优区域保留该分布的目标分布，并给出似然比解释。为此开发了采用方差最小化局部提议的序列蒙特卡洛采样器，以及结合期望与不期望 score 场 Jacobian-向量积的固定噪声优化流程。
正文
Abstract: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: this https URL
| Subjects: | Machine Learning (cs.LG) |
| Cite as: | arXiv:2610.00395 [cs.LG] |
| (or arXiv:2610.00395v1 [cs.LG] for this version) | |
| https://doi.org/10.48550/arXiv.2610.00395 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Luran Wang [view email]
[v1]
Wed, 30 Sep 2026 11:44:37 UTC (27,244 KB)
来源：arXiv：cs.LG（机器学习，全量分类） · arxiv.org