Multi-modal Manipulation Detection and Grounding on DGM4 Average across domains (test)
76.65AccuracyREFORM
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| REFORMCategory=Deepfake Detectors2026.03 | 76.65 | 65.72 | 85.17 | 60.19 | |
| AMDCategory=Deepfake Detectors2026.03 | 74.47 | 52.91 | 80.01 | 51.87 | |
| AMDTraining Domain=Guardian2025.05 | 74.47 | 52.91 | 80.01 | 51.87 | |
| FKA-OwlCategory=Deepfake Detectors2026.03 | 71.96 | 42.68 | 83.31 | 44.15 | |
| FKA-OwlTraining Domain=Guardian2025.05 | 71.96 | 42.68 | 83.31 | 44.15 | |
| HAMMER++Category=Deepfake Detectors2026.03 | 65.61 | 47.36 | 77.35 | 46.18 | |
| HAMMER++Training Domain=Guardian2025.05 | 65.61 | 47.36 | 77.34 | 46.19 | |
| HAMMERCategory=Deepfake Detectors2026.03 | 65.46 | 47.1 | 77.41 | 45.97 | |
| HAMMERTraining Domain=Guardian2025.05 | 65.45 | 47.1 | 77.41 | 45.97 | |
| LLaVa-v1.6-7BCategory=Fine-tuned LVLMs2026.03 | 65.13 | 42.61 | 67.54 | 38.58 | |
| Qwen2.5-3BCategory=Fine-tuned LVLMs2026.03 | 59.05 | 35.47 | 64 | 37.45 | |
| ViLTCategory=Deepfake Detectors2026.03 | 55.24 | 36.63 | 62.05 | 32.51 | |
| ViLTTraining Domain=Guardian2025.05 | 55.24 | 36.63 | 62.05 | 32.49 |