AI-generated image detection on AIGIBench FLUX1 (dev)
95.1AccuracyLEGO
Evaluation results
| Method | Links | ||
|---|---|---|---|
| LEGOVenue=-2026.05 | 95.1 | 98.9 | |
| DFFreqevaluation_strategy=cross-model2025.01 | 93.2 | 96.8 | |
| NPRevaluation_strategy=cross-model2025.01 | 87.2 | 93.6 | |
| AIDEevaluation_strategy=cross-model2025.01 | 84.9 | 86.6 | |
| LaDeDaevaluation_strategy=cross-model2025.01 | 84.2 | 88 | |
| FreqNetevaluation_strategy=cross-model2025.01 | 78.5 | 87.3 | |
| LGradevaluation_strategy=cross-model2025.01 | 64.6 | 70 | |
| CNN-Spotevaluation_strategy=cross-model2025.01 | 50.3 | 50.3 | |
| FatFormerevaluation_strategy=cross-model2025.01 | 48.3 | 41.8 | |
| C2P-CLIPevaluation_strategy=cross-model2025.01 | 47.6 | 46.8 | |
| UnivFDevaluation_strategy=cross-model2025.01 | 46.3 | 39.6 |