AI Paper Atlas

Research map / Image generation / Autoregressive and latent generation

All Patches Matter, More Patches Better: Enhance AI-Generated Image Detection via Panoptic Patch Learning

Zheng Yang, Ruoxin Chen, Zhiyuan Yan, Ke-Yue Zhang, Xinghe Fu, Shuang Wu, Xiujun Shu, Taiping Yao, Shouhong Ding, Zequn Qin, Xi Li

ICLR 2026 · Image generation › Autoregressive and latent generation · approach: Generalizable fake image detection

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
use margin-based Euclidean contrastive loss instead of cosine distance (ablation)PPL · GenImagemean accuracy: 92.8 -> 97.2Improved
use margin-based contrastive loss instead of InfoNCE (ablation)PPL · GenImagemean accuracy: 95.8 -> 97.2Improved
randomly select reconstructed patches instead of fixed half of the image (ablation)PPL · GenImagemean accuracy: 95.0 -> 97.2Improved
use diffusion patch replacement instead of random patch dropout (ablation)PPL · GenImagemean accuracy: 94.2 -> 97.2Improved
change reconstruction patch size from 14 to 28 (ablation)PPL · GenImagemean accuracy: 97.2 -> 97.6No clear effect
increase reconstruction patch size from 14 to 112 (ablation)PPL · GenImagemean accuracy: 97.2 -> 96.6Hurt
use diffusion reconstruction instead of stitching synthetic patches (ablation)PPL · GenImageaccuracyImproved
combine RPR and PCL rather than either alone (ablation)PPL · GenImageaccuracyImproved
set contrastive weight lambda to 0.3 (ablation)PPL · GenImageaccuracyImproved
raise reconstruction ratio to excessively high values (ablation)PPL · GenImageaccuracyHurt
vary probability of replacing fake images with RPR images (ablation)PPL · GenImageaccuracyNo clear effect
add PCL contrastive loss to CLIP-Large LoRA trainingPPL · training costGPU memory and epoch time: 16 GB -> 19 GB; 3 -> 4 hoursHurt

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