AI Paper Atlas

Research map / Image generation / Fast diffusion sampling

Compositional Sculpting of Iterative Generative Processes

Timur Garipov, Sebastiaan De Peuter, Ge Yang, Vikas Garg, Samuel Kaski, Tommi S. Jaakkola

NeurIPS 2023 · Image generation › Fast diffusion sampling · approach: Distributional control of diffusion · 0 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
compose two pre-trained GFlowNets with the harmonic mean operationGFlowNet composition · 2D gridprobability mass overlap with both base modesImproved
use contrast operation instead of harmonic meanGFlowNet composition · 2D gridmodes of p1 modulated by p2Mixed
use parameterized contrast with alpha = 0.05 (ablation)GFlowNet composition · 2D gridsuppression of p2 regionsImproved
condition on three labels instead of two (ablation)GFlowNet composition · 2D gridsharpness of compositeImproved
harmonic mean of SEH and SA GFlowNetsGFlowNet composition · molecule generation beta 32/96reward distributionImproved
contrast operation for SEH and SA GFlowNetsGFlowNet composition · molecule generation beta 32/96reward distributionMixed
condition on SEH, QED and SA togetherGFlowNet composition · molecule generation beta 32fraction of samples with all three properties highImproved
condition on two labels (SEH and QED) onlyGFlowNet composition · molecule generation beta 32reward in the unselected rewardNo clear effect
repeat the same label (SEH, SEH, SEH) when conditioningGFlowNet composition · molecule generationdistance to base distributionMixed
train classifier on terminal states first before non-terminal states (ablation)Diffusion classifier · diffusion compositionstraining noiseImproved
compose three colored-MNIST diffusion models by conditioning on y1, y2, y3Diffusion composition · colored MNISTsamples satisfying all constraintsImproved
omit conditioning on p2 in composition of y1=1, y2=1Diffusion composition · colored MNISTexcluded digitsImproved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy