D3HR maps VAE latents to a near-Gaussian space via DDIM inversion, fits per-class Gaussian statistics, and selects the most moment-matching subset to generate distilled images, reporting state-of-the-art accuracy.
Exploiting inter-sample and inter-feature relations in dataset distillation
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Taming Diffusion for Dataset Distillation with High Representativeness
D3HR maps VAE latents to a near-Gaussian space via DDIM inversion, fits per-class Gaussian statistics, and selects the most moment-matching subset to generate distilled images, reporting state-of-the-art accuracy.