AI-generated image detection on AIGIBench Mean
86.6AccuracyPGC
Evaluation results
| Method | Links | ||
|---|---|---|---|
| PGCTraining Data=ProGAN (4 classes) and SDv1.42026.05 | 86.6 | 95 | |
| DDATraining Data=ProGAN (4 classes) and SDv1.42026.05 | 83.1 | 89.9 | |
| LEGOVenue=-2026.05 | 80.2 | 89.6 | |
| SAFETraining Data=ProGAN (4 classes) and SDv1.42026.05 | 80 | 82.6 | |
| DFFreqevaluation_strategy=cross-model2025.01 | 77.9 | 82 | |
| AIDETraining Data=ProGAN (4 classes) and SDv1.42026.05 | 77.6 | 82.5 | |
| AIDEVenue=ICLR’252026.05 | 76.8 | 81.8 | |
| AIDEevaluation_strategy=cross-model2025.01 | 74.9 | 76.2 | |
| EffortVenue=ICML’252026.05 | 74.1 | 80.2 | |
| LaDeDaVenue=arXiv’242026.05 | 72.6 | 78.6 | |
| UnivFDTraining Data=ProGAN (4 classes) and SDv1.42026.05 | 72.5 | 75.6 | |
| CO-SPYVenue=CVPR’252026.05 | 71.5 | 78.2 | |
| CLIPDetectionVenue=CVPR’232026.05 | 71.4 | 74.6 | |
| C2P-CLIPevaluation_strategy=cross-model2025.01 | 71 | 77.3 | |
| B-FreeTraining Data=ProGAN (4 classes) and SDv1.42026.05 | 70.6 | 82.5 | |
| DFFreqVenue=arXiv’252026.05 | 70 | 74.7 | |
| FatFormerevaluation_strategy=cross-model2025.01 | 69.9 | 74.3 | |
| LaDeDaevaluation_strategy=cross-model2025.01 | 68.8 | 71.3 | |
| NPRevaluation_strategy=cross-model2025.01 | 68.2 | 70.8 | |
| NPRTraining Data=ProGAN (4 classes) and SDv1.42026.05 | 67.9 | 73.5 | |
| NPRVenue=CVPR’242026.05 | 66.6 | 72.4 | |
| FreqNetTraining Data=ProGAN (4 classes) and SDv1.42026.05 | 66.1 | 70.1 | |
| FreqNetVenue=AAAI’232026.05 | 64.8 | 68.9 | |
| UnivFDevaluation_strategy=cross-model2025.01 | 63.3 | 66.4 | |
| LGradTraining Data=ProGAN (4 classes) and SDv1.42026.05 | 62.9 | 66.6 | |
| LGradVenue=CVPR’232026.05 | 61.5 | 65.2 | |
| ResNet-50Venue=CVPR’162026.05 | 60.3 | 68 | |
| FreqNetevaluation_strategy=cross-model2025.01 | 60.2 | 61.7 | |
| Gram-netVenue=CVPR’202026.05 | 56.9 | 60.8 | |
| CNNDetTraining Data=ProGAN (4 classes) and SDv1.42026.05 | 55.1 | 67.1 | |
| CNN-Spotevaluation_strategy=cross-model2025.01 | 53.6 | 57.6 | |
| LGradevaluation_strategy=cross-model2025.01 | 53.4 | 56.2 | |
| CNNDetectionVenue=CVPR’202026.05 | 53.4 | 65.7 |