AI Paper Atlas

Research map / Image generation / Fast diffusion sampling

Gradient Preconditioning for Efficient and Reliable Reward-Guided Generation

Jisung Hwang, Minhyuk Sung

ICML 2026 · 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
project reward gradients onto the white Gaussian noise feasible set instead of regularizing the latenttest-time latent optimization with FLUX-schnell · Aesthetic Score optimization, 200 iterationsAesthetic Score (target): 7.133 -> 8.908 (vs MPGR)Improved
use gradient preconditioning with only 60 iterationsgradient preconditioning · Aesthetic Score optimization, FLUXwall-clock time (s) at matched Aesthetic Score: 235.5 -> 69.7Improved
use gradient preconditioning instead of regularizationFLUX-schnell · held-out PickScore under Aesthetic Score optimizationPickScore (held-out): 0.219 -> 0.220No clear effect
add gradient preconditioning versus unoptimized samplingFLUX-schnell · diversity under Aesthetic Score optimizationInception Score / Vendi Score: IS 22.33 -> 21.10; Vendi 6.42 -> 6.97No clear effect
add projection step to every iterationgradient preconditioning · FLUX runtimeshare of runtime: 0.04% of runtimeNo clear effect
tune the regularization weight lambda for baselines (ablation)regularization-based baselines · reward-guided generation, FLUXreward trade-offNo clear effect
use soft spatial-correlation regularizers (PRNO, MPGR) to avoid reward hackingbaseline regularizers · qualitative FLUX resultsrealism of outputsMixed

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