AI Paper Atlas

Research map / Image generation / Autoregressive and latent generation

A Difference-in-Difference Approach to Detecting AI-Generated Images

Xinyi Qi, Kai Ye, Chengchun Shi, Ying Yang, Jin Zhu, Hongyi Zhou

CVPR 2026 · Image generation › Autoregressive and latent generation · approach: Reconstruction-based diffusion image detection · 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
add second-order difference to first-order DIREDIRE · Trained on LAION & Kan.3 (total avg)accuracy (%): 92.96 -> 94.55Improved
add second-order difference to first-order DIREDIRE · Trained on LAION & SDXL (total avg)accuracy (%): 94.83 -> 96.42Improved
add second-order difference with large aligned training setDIRE · Trained on ImageNet & ADM (total avg)accuracy (%): 99.37 -> 99.41No clear effect
add second-order difference for transfer to GAN imagesDIRE · GAN-generated images (total avg)accuracy (%): 92.94 -> 94.42Improved
add first-order difference to second-order only (ablation)Delta2 · Ablation LSUN-B ADMaccuracy (%): 69.0 -> 86.5Improved
add first-order difference to second-order only (ablation)Delta2 · Ablation LSUN-B PNDMaccuracy (%): 81.5 -> 83.0Improved
add first-order difference to second-order only (ablation)Delta2 · Ablation LAION Kan.3accuracy (%): 100 -> 99.9No clear effect
use second-order only instead of DIRE first-order (ablation)DIRE · Ablation LSUN-B PNDMaccuracy (%): 67.5 -> 81.5Improved
use second-order only instead of DIRE first-order (ablation)DIRE · Ablation LSUN-B ADMaccuracy (%): 84.0 -> 69.0Hurt
add second-order only instead of DIRE first-order (ablation)DIRE · Ablation ImageNet ADMaccuracy (%): 91.0 -> 98.4Improved
test on JPEG75 instead of PNGDID · Trained on LAION & Kan.3 PNG-only (avg)accuracy (%): 94.09 -> 83.15Hurt

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

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