Research map / Image generation / Autoregressive and latent generation
PGC: Peak-Guided Calibration for Generalizable AI-Generated Image Detection
Xiaoyu Zhou, Jianwei Fei, Peipeng Yu, Jingchang Xie, Chong Cheng, Zhihua Xia
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 PGCM on top of residual+RGB concatenation (ablation) | PGC · GenImage | mean accuracy: 92.1 -> 98.3 | Improved |
| add RGB stream to residual stream via simple concatenation (ablation) | residual-only baseline · GenImage | mean accuracy: 93.4 -> 92.1 | Hurt |
| add RGB stream to residual stream (needed for BigGAN) (ablation) | PGC · GenImage BigGAN | accuracy: 50.0 -> 90.9 | Improved |
| scale backbone from DINOv2-Base to Large (ablation) | PGC · GenImage | mean accuracy: 95.8 -> 98.3 | Improved |
| scale backbone from DINOv2-Large to Giant (ablation) | PGC · GenImage | mean accuracy: 98.3 -> 97.9 | Hurt |
| PGC versus previous best CoD | GenImage detectors · GenImage | mean accuracy: 96.2 -> 98.3 | Improved |
| PGC versus nearest competitor DDA | CommGen15 detectors · CommGen15 | overall accuracy: +12.3% | Improved |
| PGC versus strongest baseline DDA | AIGI detectors · AIGI | mean accuracy: +3.5% in Acc and +5.1% in AP | Improved |
| PGC versus previous best AIDE | UniversalFakeDetect detectors · UniversalFakeDetect | mean AP: +2.2% | Improved |
| increase transfer PGD perturbation budget from 0.3 to 0.5 against PGC | PGC · CommGen15 adversarial | mean accuracy: 98.2 -> 98.1 -> 95.0 -> 89.5 | Hurt |
Closest papers by meaning
- Beyond Semantic Features: Pixel-level Mapping for Generalized AI-Generated Image Detection (AAAI 2026 · similarity 0.86)
- Your AI-Generated Image Detector Can Secretly Achieve SOTA Accuracy, If Calibrated (AAAI 2026 · similarity 0.85)
- Is Artificial Intelligence Generated Image Detection a Solved Problem? (NeurIPS 2025 · similarity 0.85)
- CO-SPY: Combining Semantic and Pixel Features to Detect Synthetic Images by AI (CVPR 2025 · similarity 0.84)
- ReCon: Region-Controllable Data Augmentation with Rectification and Alignment for Object Detection (NeurIPS 2025 · similarity 0.84)
- Direct Discriminative Optimization: Your Likelihood-Based Visual Generative Model is Secretly a GAN Discriminator (ICML 2025 · similarity 0.84)
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.
