Abstract:While diffusion models excel at generating high-quality images, their tendency to memorize training data poses significant privacy and copyright risks. In this work, we for the first time identify that memorization induces internal numerical instability, often manifesting as visually ``broken'' artifacts. Inspired by stability analysis in numerical methods, we introduce empirical stability regions based on latent update norms to quantitatively characterize stable behavior during generation. Leveraging this, we propose a principled, on-the-fly framework for step-wise detection and adaptive mitigation. Our approach suppresses memorization without altering prompts or guidance, thereby preserving semantic fidelity and image quality. Extensive experiments on Stable Diffusion 1.4 demonstrate that our method achieves an AUC $>0.999$ detection performance and a $0.0\%$ memorization rate after mitigation with negligible overhead ($\approx0.01$s per image).
| Comments: | KDD 2026, extended version |
| Subjects: | Computer Vision and Pattern Recognition (cs.CV) |
| Cite as: | arXiv:2605.22050 [cs.CV] |
| (or arXiv:2605.22050v2 [cs.CV] for this version) | |
| https://doi.org/10.48550/arXiv.2605.22050 arXiv-issued DOI via DataCite |
|
| Related DOI: | https://doi.org/10.1145/3770855.3817770
DOI(s) linking to related resources |
From: Yuanmin Huang [view email]
[v1]
Thu, 21 May 2026 06:36:59 UTC (23,812 KB)
[v2]
Fri, 22 May 2026 16:38:43 UTC (23,830 KB)
此内容由惯性聚合(RSS阅读器)自动聚合整理，仅供阅读参考。 原文来自 — 版权归原作者所有。