AI Paper Atlas

Research map / Image generation / Text-to-image diffusion

It's Never Too Late: Noise Optimization for Collapse Recovery in Trained Diffusion Models

Anne Harrington, A. Sophia Koepke, Shyamgopal Karthik, Trevor Darrell, Alexei A. Efros

CVPR 2026 · Image generation › Text-to-image diffusion · approach: Inference-time T2I improvement · 7 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
optimize initial noise to maximize diversity (DINO+CLIP) instead of i.i.d. noiseSDXL-Turbo · GenEval, white noiseDINO diversity: 0.588 -> 0.784Improved
noise optimization versus Parmar et al. candidate selectionSDXL-Turbo · GenEval, white noiseDINO diversity: 0.705 -> 0.784Improved
optimize noise with DPP diversity and HPSv2 rewardSDXL-Turbo · GenEvalVendi score: 2.000 -> 4.000Improved
optimize noise with DPP diversity and HPSv2 rewardSDXL-Turbo · GenEvalHPSv2: 0.284 -> 0.292No clear effect
optimize noise with DPP and HPSv2Flux.1 [schnell] · GenEval, white noiseVendi score: 2.013 -> 2.753Improved
optimize noise with DPP and HPSv2Flux.1 [schnell] · GenEval, white noiseHPSv2: 0.304 -> 0.293Hurt
use set-level objectives (DPP, Vendi) instead of pairwise DINO objective (ablation)noise optimization · SDXL-Turbo human preference studywin rateImproved
pink noise (alpha=0.2) initialization instead of white (ablation)i.i.d. sampling, SDXL-Turbo · GenEvalDINO diversity: 0.588 -> 0.642Improved
pink noise (alpha=0.2) initialization instead of white (ablation)noise optimization, SDXL-Turbo · GenEvalDINO diversity: 0.784 -> 0.786Improved
pink noise initialization for Parmar et al. (ablation)SDXL-Turbo · GenEvalDINO diversity: 0.705 -> 0.749Improved
increase pink noise exponent alpha above 0.2 (ablation)noise initialization · GenEvalimage qualityHurt
pink noise initializationnoise optimization · GenEvaliterations to beat Parmar et al.: 9 -> 12 iterations (pool 64, white vs alpha=0.2 pink)Hurt

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