arXiv：cs.LG· David Dinkevich, Nisan Chiprut, Yoav HaCohen, Dani Lischinski·· 13 小时前
APEX：面向极端压缩的异步前缀去噪
APEX: Asynchronous Prefix Denoising for Extreme Compression
AI 导读
APEX 是一种面向极端潜在压缩的异步前缀去噪方法，通过分层潜在空间与按通道组错开局部时间的去噪，在不增加 token 的前提下缓解压缩带来的重建-生成矛盾。在 ImageNet-512 上，APEX 在 64x 和 128x 压缩下所有评测通道数均优于现有方法，并在等训练与采样算力下于各预算档位超过 256-token 模型，低算力时增益更大。代码已开源。
正文
Abstract:Fewer latent tokens promise cheaper diffusion. Yet aggressive latent compression creates a reconstruction-generation dilemma: adding channels improves reconstruction but makes generation harder. We introduce APEX, short for Asynchronous Prefix denoising for EXtreme compression, which pairs a hierarchical latent space with asynchronous denoising over channel groups. Our autoencoder orders channel prefixes from structure to detail, while one shared denoiser generates them at staggered local times without adding tokens. On ImageNet-512, APEX outperforms existing methods at every evaluated channel count under 64x and 128x compression. At matched training and guided sampling compute, it also outperforms a 256-token model at every evaluated budget, with larger gains at lower compute. Code is available at this https URL.
| Comments: | 24 pages |
| Subjects: | Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG) |
| Cite as: | arXiv:2610.12843 [cs.CV] |
| (or arXiv:2610.12843v1 [cs.CV] for this version) | |
| https://doi.org/10.48550/arXiv.2610.12843 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: David Dinkevich [view email]
[v1]
Fri, 9 Oct 2026 01:06:57 UTC (39,279 KB)
来源：arXiv：cs.LG · arxiv.org