SLCD provably converges, under no-regret learning and a strong score-estimation assumption, to the KL-regularized optimal distribution using only supervised classification oracles.
Improved analysis of score-based generative modeling: User-friendly bounds under minimal smoothness assumptions
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
-
Efficient Controllable Diffusion via Optimal Classifier Guidance
SLCD provably converges, under no-regret learning and a strong score-estimation assumption, to the KL-regularized optimal distribution using only supervised classification oracles.