Trust Sampling improves training-free constrained diffusion by allowing multiple normalized gradient steps per denoising timestep, with a variance-based trust schedule and a predicted-noise manifold boundary for early stopping.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2024 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Constrained Diffusion with Trust Sampling
Trust Sampling improves training-free constrained diffusion by allowing multiple normalized gradient steps per denoising timestep, with a variance-based trust schedule and a predicted-noise manifold boundary for early stopping.