AI Paper Atlas

Research map / Image generation / Fast diffusion sampling

Offline Preference Optimization for Rectified Flow with Noise-Tracked Pairs

Yunhong Lu, Qichao Wang, Hengyuan Cao, Xiaoyin Xu, Min Zhang

ICML 2026 · Image generation › Fast diffusion sampling · approach: Flow matching distillation · 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
add stored prior noise and noise-image interpolation to DPO (ablation)Diffusion-DPO on FLUX.1-dev · HPDv2HPSv2.1: 30.84 -> 31.08Improved
add dynamic regularization on top of prior noise (ablation)PNAPO on FLUX.1-dev · HPDv2HPSv2.1: 31.08 -> 31.71Improved
add dynamic regularization without prior noise (ablation)Diffusion-DPO on FLUX.1-dev · HPDv2HPSv2.1: 30.84 -> 30.96Improved
add reward-gap controller f(delta r) to beta (ablation)PNAPO · HPDv2HPSv2.1: 31.08 -> 31.66Improved
add training-progress annealing g(n) to beta (ablation)PNAPO · HPDv2HPSv2.1: 31.08 -> 31.13Improved
use HPSv2.1 as reward model instead of ImageReward (ablation)PNAPO · HPDv2HPSv2.1: 31.10 -> 31.71Improved
use Aesthetic reward model instead of HPSv2.1 (ablation)PNAPO · HPDv2Aesthetic: 6.475 -> 6.509 (Aesthetic), CLIP 34.71 -> 34.57Mixed
lower first threshold n1 from 1000 to 500 (ablation)dynamic regularization schedule · HPDv2HPSv2.1: 31.71 -> 30.96Hurt
increase n2 from 2000 to 4000 (ablation)dynamic regularization schedule · HPDv2HPSv2.1: 31.71 -> 31.62No clear effect
train with PNAPO instead of Diffusion-DPOFLUX.1-dev · training costGPU-hours: ~422.4 -> ~35.2Improved
apply PNAPO to SD3-MSD3-M · GenEvaloverall: 0.68 -> 0.73Improved
apply PNAPO to FLUX on counting attributeFLUX.1-dev · GenEvaloverall: 0.65 -> 0.69Improved

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