Diffusion model features, when decoded with k-sparse autoencoders, reveal interpretable visual concepts, and a lightweight classifier on the best layer (up_ft1 at t=25) beats prior diffusion-based classifiers on fine-grained benchmarks.
https://www.midjourney.com/home
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2024 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
$\textit{Revelio}$: Interpreting and leveraging semantic information in diffusion models
Diffusion model features, when decoded with k-sparse autoencoders, reveal interpretable visual concepts, and a lightweight classifier on the best layer (up_ft1 at t=25) beats prior diffusion-based classifiers on fine-grained benchmarks.