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
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 changed | Where | Measured | Effect |
|---|---|---|---|
| add second-order difference to first-order DIRE | DIRE · Trained on LAION & Kan.3 (total avg) | accuracy (%): 92.96 -> 94.55 | Improved |
| add second-order difference to first-order DIRE | DIRE · Trained on LAION & SDXL (total avg) | accuracy (%): 94.83 -> 96.42 | Improved |
| add second-order difference with large aligned training set | DIRE · Trained on ImageNet & ADM (total avg) | accuracy (%): 99.37 -> 99.41 | No clear effect |
| add second-order difference for transfer to GAN images | DIRE · GAN-generated images (total avg) | accuracy (%): 92.94 -> 94.42 | Improved |
| add first-order difference to second-order only (ablation) | Delta2 · Ablation LSUN-B ADM | accuracy (%): 69.0 -> 86.5 | Improved |
| add first-order difference to second-order only (ablation) | Delta2 · Ablation LSUN-B PNDM | accuracy (%): 81.5 -> 83.0 | Improved |
| add first-order difference to second-order only (ablation) | Delta2 · Ablation LAION Kan.3 | accuracy (%): 100 -> 99.9 | No clear effect |
| use second-order only instead of DIRE first-order (ablation) | DIRE · Ablation LSUN-B PNDM | accuracy (%): 67.5 -> 81.5 | Improved |
| use second-order only instead of DIRE first-order (ablation) | DIRE · Ablation LSUN-B ADM | accuracy (%): 84.0 -> 69.0 | Hurt |
| add second-order only instead of DIRE first-order (ablation) | DIRE · Ablation ImageNet ADM | accuracy (%): 91.0 -> 98.4 | Improved |
| test on JPEG75 instead of PNG | DID · Trained on LAION & Kan.3 PNG-only (avg) | accuracy (%): 94.09 -> 83.15 | Hurt |
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.
- AEROBLADE — AEROBLADE: Training-Free Detection of Latent Diffusion Images Using Autoencoder Reconstruction Error (CVPR 2024) · better on 2 of 2 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
- Beyond Generation: A Diffusion-based Low-level Feature Extractor for Detecting AI-generated Images (CVPR 2025 · similarity 0.89)
- Aligned Datasets Improve Detection of Latent Diffusion-Generated Images (ICLR 2025 · similarity 0.87)
- Difference Inversion: Interpolate and Isolate the Difference with Token Consistency for Image Analogy Generation (CVPR 2025 · similarity 0.87)
- Stay-Positive: A Case for Ignoring Real Image Features in Fake Image Detection (ICML 2025 · similarity 0.87)
- Breaking Semantic Artifacts for Generalized AI-generated Image Detection (NeurIPS 2024 · similarity 0.87)
- CO-SPY: Combining Semantic and Pixel Features to Detect Synthetic Images by AI (CVPR 2025 · similarity 0.87)
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.
