arXiv cs.AI· Hongxu Chen, Chun Wang, Yanghao Wang, Hongxiang Li, Zhen Wang, Zhixiang Zhang, Long Chen·· 10 小时前AI 评分31
DM-VAE:用双流形结构约束 VAE 潜空间
Dual-Manifold Variational Autoencoders
AI 导读
研究者提出双流形 VAE(DM-VAE),用高维重建分支搭配低维分支提取干净的拓扑骨架,将高维潜流形锚定到更平滑的子空间,无需依赖 DINOv2 等外部基础模型。在 ImageNet-1K(256x256)上,该无基础模型方案取得 SOTA 的 rFID 0.12,生成 FID50K 为 1.28(无 CFG)和 1.20(有 CFG),与 DINOv2 正交结合后进一步达到 1.11。
正文 · 原文
Abstract:In latent diffusion models, expanding the latent dimension of variational autoencoders (VAEs) is essential to push the upper bound of reconstruction quality. However, without explicit structural constraints, such unregularized expansion easily yields a rugged and ill-conditioned latent manifold, causing generated latents to drift off the manifold during generation (i.e., off-manifold problem). Recent works smooth this manifold by using external foundation models (e.g., DINOv2) as a topological regularizer. However, we argue that tying the latent space to frozen external priors inherently restricts generalization: bound by their pre-training distributions and semantic biases, they may fail to capture faithful visual details on out-of-distribution domains. Observing that low-dimensional latents naturally form a smoother, generation-friendly manifold, we propose dual-manifold VAE (DM-VAE) to leverage internal representations as a topological regularizer. Specifically, DM-VAE pairs a high-dimensional reconstructive branch with a low-dimensional branch that extracts a clean topological skeleton, anchoring the high-dimensional manifold to this smooth subspace. Building upon this framework, we further introduce topological representation alignment (TRA) to supervise intermediate generator features during training, and internal guidance with topological skeleton (IG_TS) to guide structural components while preserving high-frequency details at inference. On ImageNet-1K (256x256), our foundation-model-free DM-VAE achieves a state-of-the-art rFID of 0.12 and a generation FID50K of 1.28 (w/o CFG) and 1.20 (w/CFG), and further reaches 1.11 when orthogonally combined with DINOv2.
| Subjects: | Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2611.23964 [cs.CV] |
| (or arXiv:2611.23964v1 [cs.CV] for this version) | |
| https://doi.org/10.48550/arXiv.2611.23964 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Hongxu Chen [view email]
[v1]
Wed, 18 Nov 2026 06:58:42 UTC (8,984 KB)
来源:arXiv cs.AI · arxiv.org