AI Paper Atlas

Research map / Image generation / Fast diffusion sampling

From Predictors to Samplers via the Training Trajectory

Soumya Ram, Akhila Ram

ICLR 2026 · Image generation › Fast diffusion sampling · approach: Diffusion theory and design space

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
anneal across training checkpoints instead of final checkpoint only (ablation)GWG sampling · Boolean polynomial targetsuccess rate: 0.04 -> 0.5200Improved
anneal across training checkpoints instead of final checkpoint only (ablation)GWG sampling · Indicator d=10success probability: 0.12 -> 0.99Improved
increase indicator size d from 3 to 10 (ablation)GWG on final checkpoint · Indicator functionsuccess probability: 0.47 -> 0.12Hurt
increase indicator size d from 3 to 10 (ablation)trajectory annealing · Indicator functionsuccess probability: 0.98 -> 0.99No clear effect
anneal across training checkpoints with adversarial linear term (ablation)GWG sampling · Non-convex binary landscapesuccess rate: 0.080 -> 1.00Improved
anneal across training checkpoints with three length-5 indicators (ablation)GWG sampling · 3 length-5 indicatorssuccess rate: 0.0250 -> 1.00Improved
anneal across training checkpoints (1K steps) (ablation)Temp-GWG · Binary MNIST EBMFID: 29.61 -> 11.73Improved
anneal across training checkpoints (10K steps) (ablation)Temp-GWG · Binary MNIST EBMFID: 21.12 -> 5.49Improved
use 5 checkpoints instead of baseline final-checkpoint annealing (ablation)trajectory annealing · Binary MNIST EBM, 1K stepsFID: 29.61 -> 16.10Improved
increase number of checkpoints from 5 to 500 (ablation)trajectory annealing · Binary MNIST EBM, 1K stepsFID: 16.10 -> 11.73Improved
anneal across training checkpoints (ablation)GWG sampling · DNA design (MAX TF)median fitness: 2.72 -> 10.04Improved
anneal across training checkpoints under Hamming-7 constraint (ablation)GWG sampling · Constrained DNA designmotif (%): 31.3 -> 63.0Improved

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