Multi-modal Forgery Detection and Grounding on MDSM NYT
94.67AccuracyFKA-Owl
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| FKA-OwlSetting=Tr. on NYT2025.05 | 94.67 | 78.18 | 55.81 | |
| AMDSetting=Tr. on NYT2025.05 | 92.24 | 84.47 | 72.94 | |
| AMDSetting=Tr. on Guardian2025.05 | 84.29 | 48.54 | 52.38 | |
| ViLTSetting=Tr. on NYT2025.05 | 83.27 | 64.27 | 22.73 | |
| FKA-OwlSetting=Tr. on Guardian2025.05 | 80.6 | 40.44 | 26.33 | |
| HAMMER++Setting=Tr. on NYT2025.05 | 79.61 | 57.06 | 54.44 | |
| HAMMERSetting=Tr. on NYT2025.05 | 79.2 | 55.86 | 51.34 | |
| ViLTSetting=Tr. on Guardian2025.05 | 68.8 | 43.99 | 21.77 | |
| HAMMER++Setting=Tr. on Guardian2025.05 | 62.75 | 36.45 | 23.76 | |
| HAMMERSetting=Tr. on Guardian2025.05 | 61.89 | 37.98 | 18.84 | |
| Gemini-2.0Setting=Zero-Shot2025.05 | 56.05 | 33.16 | 1.44 | |
| GPT-4oSetting=Zero-Shot2025.05 | 48.48 | 27.9 | 0.82 | |
| Qwen2.5-VLSetting=Zero-Shot2025.05 | 47.74 | 29.24 | 0 | |
| Qwen2-a22bSetting=Zero-Shot2025.05 | 45.29 | 25.01 | 0.71 |