AI Paper Atlas

Research map / Image generation / Autoregressive and latent generation

Beyond Artifacts: Real-Centric Envelope Modeling for Reliable AI-Generated Image Detection

Ruiqi Liu, Yi Han, Zhengbo Zhang, Liwei Yao, Zhiyuan Yan, Jialiang Shen, ZhiJin Chen, Manni Cui, Boyi Sun, Lubin Weng, Jing Dong, Yan Wang and 1 more

ECCV 2026 · Image generation › Autoregressive and latent generation · approach: Generalizable fake image detection · 17 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
use feature-level perturbation in MBR instead of plain VAE reconstruction (ablation)REM · RealChain (Chain Degradations)B.Acc: 3.4% drop without itImproved
add tangency regularization to the envelope estimator (ablation)REM · RealChain (Chain Degradations)B.Acc: 2.7% drop without itImproved
add cross-domain consistency (anchor and residual) with degradation augmentation (ablation)REM · RealChain (Chain Degradations)B.Acc: 16.4% drop without itImproved
train on DDA-aligned data instead of GenImage (ablation)AIDE · Seven-benchmark averageAvg B.Acc: 68.0 -> 72.7Improved
train on DDA-aligned data instead of GenImage (ablation)DINOv3 · Seven-benchmark averageAvg B.Acc: 87.4 -> 90.3Improved
train with REM instead of DDA data (ablation)DINOv3 · Seven-benchmark averageAvg B.Acc: 90.3 -> 93.5Improved
scale up DINO backbone (v3-L/16 to v3-7B/16) (ablation)REM · RealChain (Chain Degradations)B.Acc: 83.3 -> 86.1Improved
use DINOv3 instead of DINOv2 backbone (L vs L) (ablation)REM · RealChain (Chain Degradations)B.Acc: 78.5 -> 83.3Improved
increase number of real training samples up to 100k (ablation)REM · all benchmarksaccuracyImproved
replace artifact-driven detectors with REMREM vs DDA · RealChain (Chain Degradations)B.Acc: 65.8 -> 84.2Improved
REM vs DDA on in-the-wild benchmarksREM · In-the-wild (4 datasets)Avg B.Acc: 88.3 -> 94.8Improved
REM vs DDA on source attributionREM · AIGCDetect source attributionaccuracy: 68.0 -> 75.7Improved

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