Unverified50% confidenceFactExact time
Stable Diffusion系列使用KL-regularized VAE，通过正则化约束潜空间的平滑性来缓解误差放大，但不能完全消除放大效应
1
Sources
50%
Confidence
Long-term
Relevance
8/8/2026
First Seen
Sources
Related Claims
UnverifiedVAE的损失函数由重建损失（常用MSE或BCE）和KL散度两部分构成，KL散度约束潜在分布接近标准正态分布起正则化作用68% similarUnverifiedStability AI开源的Latent Diffusion Model将扩散过程转移到低维潜空间，使计算成本下降一个数量级，催生了Stable Diffusion64% similarVerifiedStable Diffusion performs the denoising diffusion process in a compressed latent space rather than pixel space, significantly reducing computational resource requirements.62% similarUnverifiedFlow Matching 通过学习连续概率流以更少的推理步骤实现同等质量生成，相比 Diffusion 显著降低延迟60% similarUnverified扩散模型（Diffusion Model）的兴起使Stable Diffusion、Midjourney等工具将图像生成门槛降至消费级硬件可运行水平60% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/707121API
curl https://kongchang.com/api/v1/knowledge/claims/707121MCP
get_claim(id=707121)