AI Paper Atlas

Research map / Image generation / Fast diffusion sampling

Neon: Negative Extrapolation From Self-Training Improves Image Generation

Sina Alemohammad, Zhangyang Wang, Richard Baraniuk

ICLR 2026 · Image generation › Fast diffusion sampling · approach: Distributional control of diffusion · 3 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
apply Neon negative extrapolation to self-trained EDM-VPEDM-VP · CIFAR-10 conditionalFID: 1.78 -> 1.38Improved
apply Neon to EDM-VPEDM-VP · FFHQ-64FID: 2.39 -> 1.12Improved
apply Neon to flow matchingFlow matching · CIFAR-10FID: 3.5 -> 2.32Improved
apply Neon to xAR-LxAR-L · ImageNet-256FID: 1.28 -> 1.02Improved
apply Neon to VAR-d16VAR-d16 · ImageNet-256FID: 3.30 -> 2.01Improved
apply Neon to IMM at 1 stepIMM · ImageNet-256 (T=1)FID: 7.77 -> 6.67Improved
jointly tune merge weight w and CFG scale gamma (ablation)Neon + VAR-d16 · ImageNet-256FID: 3.01 -> 2.01Improved
use synthetic data from other architectures (flow, IMM) for Neon on EDM-VP (ablation)EDM-VP · CIFAR-10FID: 1.97 -> 1.59 (flow); 1.97 -> 1.80 (IMM); 1.97 -> 1.38 (self)Improved
replace synthetic data with CIFAR-10C corrupted real images (ablation)Neon + EDM-VP · CIFAR-10FIDNo clear effect
use much smaller synthetic dataset (1k samples) for xAR (ablation)Neon + xAR-L · ImageNet-256FID: 1.05 vs 1.02 at 750kNo clear effect
use fewer than 90k synthetic samples for VAR-d16 (ablation)Neon + VAR-d16 · ImageNet-256FIDHurt
raise CFG scale used to generate synthetic data to extreme values (>=6) (ablation)Neon + xAR-B · ImageNet-256FID: 1.30-1.31 for gamma in [1,3] then degradesHurt

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