Unverified50% confidenceFactExact time
The VAE used by Stable Diffusion has been specifically optimized with perceptual loss and adversarial training, and its reconstruction quality far surpasses that of early standard VAEs, but issues such as high-frequency detail loss and color shift still remain
1
Sources
50%
Confidence
Long-term
Relevance
10/9/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedStable Diffusion通过预训练的VAE将512×512×3的图像编码为64×64×4的潜在张量，计算量降低约48倍66% similarUnverifiedNormalizing flows stand in contrast to VAEs (which require variational lower-bound approximation) and GANs (which cannot compute likelihoods)63% similarVerifiedStable Diffusion compresses the diffusion process from pixel space to latent space, implemented by a Variational Autoencoder (VAE).62% similarUnverifiedRQ-VAE progressively approximates the original vector through multi-level residuals, with the first quantization layer capturing coarse-grained semantics and subsequent layers quantizing residuals to capture finer-grained features.61% similarVerifiedOverfitting refers to the phenomenon where a model performs excellently on training data but its performance drops significantly on unseen test data.60% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/992648API
curl https://kongchang.com/api/v1/knowledge/claims/992648MCP
get_claim(id=992648)