1 paper
Aidan Gleich, Scott C. Schmidler
We address the problem of accurate, training-free guidance for conditional generation in trained diffusion models. Existing methods typically rely on point-estimates to approximate…