AI Paper Atlas

Research map / Image generation / Fast diffusion sampling

Generalization and Memorization in Rectified Flow

Mingxing Rao, Daniel Moyer

ECCV 2026 · Image generation › Fast diffusion sampling · approach: Rectified flow straightening · 2 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
average N Monte Carlo noise samples (T_mc, #mc=5) instead of single-noise T_naivemembership inference on RF · CIFAR-10AUC: 70.98 -> 75.12Improved
T_mc instead of T_naivemembership inference on RF · SVHNTPR@1%FPR: 1.89 -> 1.54Hurt
add input complexity calibration (divide by JPEG size) to T_mc (ablation)T_mc · CIFAR-10TPR@1%FPR: 3.05 -> 27.88Improved
add input complexity calibration to T_mc (ablation)T_mc · TinyImageNetTPR@1%FPR: 5.33 -> 50.03Improved
add input complexity calibration to T_mc (ablation)T_mc · SVHNAUC: 70.92 -> 79.43Improved
add complexity calibration to T_naive (ablation)T_naive · TinyImageNetAUC: 72.33 -> 79.61Improved
add complexity calibration to T_naive (ablation)T_naive · CIFAR-10TPR@1%FPR: 2.70 -> 3.70No clear effect
use JPEG instead of PNG for complexity measure (ablation)T_mc_cal · CIFAR-10TPR: 11.84 -> 23.35Improved
use T_mc instead of T_naive across timestepsmembership inference · MSCOCO t=0.5AUC: 76.14 -> 85.33Improved
larger Monte Carlo sample count N_mc from 1 (ablation)T_mc · CIFAR-10AUCImproved
use Symmetric Exponential timestep sampling (alpha=2, 4) instead of uniformRectified Flow training · CIFAR-10, SVHNpeak MIA AUC and FIDImproved
non-identity data covariance (condition number kappa 1.0 to 100.0)Rectified Flow toy Gaussian · Toy Gaussianpeak t: 0.50 -> 0.19Hurt

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