Auto-regressive diffusion models provably control conditional-distribution sampling error with only a factor-K increase in inference cost, unlike vanilla diffusion where conditional error can blow up despite small joint error.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Capturing Conditional Dependence via Auto-regressive Diffusion Models
Auto-regressive diffusion models provably control conditional-distribution sampling error with only a factor-K increase in inference cost, unlike vanilla diffusion where conditional error can blow up despite small joint error.