AI-generated video detection on GenVideo Gen2
97.2AccuracyDeMamba
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| DeMambaTraining=10K Kinetics-400 + 10K Pika2026.05 | 97.2 | 98 | 97.22 | 99.46 | |
| DeMambaTraining Data=Kinetics-400 and Pika, Evaluation Protocol=Standard2025.10 | 97.2 | 98 | 97.22 | 99.46 | |
| TALLTraining=10K Kinetics-400 + 10K Pika2026.05 | 96.9 | 94.8 | 96.83 | 99.79 | |
| TALLTraining Data=Kinetics-400 and Pika, Evaluation Protocol=Standard2025.10 | 96.9 | 94.8 | 96.83 | 99.79 | |
| Skyra-RL-GenVideoModel Size=7B2025.12 | 95.25 | 96.23 | 95.3 | — | |
| NPRTraining=10K Kinetics-400 + 10K Pika2026.05 | 94.8 | 91.2 | 94.61 | 99.13 | |
| NPRTraining Data=Kinetics-400 and Pika, Evaluation Protocol=Standard2025.10 | 94.8 | 91.2 | 94.61 | 99.13 | |
| CAM-VFDTraining=10K Kinetics-400 + 10K Pika2026.05 | 93.2 | 86.7 | 89.9 | 96.4 | |
| DeMambaTraining Data=Kinetics-400 and SEINE2025.10 | 91.9 | 85.6 | 91.36 | 98.76 | |
| TALLTraining Data=Kinetics-400 and SEINE2025.10 | 88.4 | 77.8 | 87.02 | 97.76 | |
| NSG-VDTraining=10K Kinetics-400 + 10K Pika2026.05 | 88.33 | 80 | 87.27 | 94.83 | |
| NSG-VDTraining Data=Kinetics-400 and Pika, Evaluation Protocol=Standard2025.10 | 88.33 | 80 | 87.27 | 94.83 | |
| NSG-VDTraining Data=Kinetics-400 and SEINE2025.10 | 87.08 | 98.33 | 88.39 | 97.87 | |
| NPRTraining Data=Kinetics-400 and SEINE2025.10 | 85.7 | 75 | 83.99 | 95.13 | |
| STILTraining Data=Kinetics-400 and SEINE2025.10 | 83.1 | 66.4 | 79.71 | 98.06 | |
| Skyra-RLModel Size=7B2025.12 | 75.94 | 54.49 | 69.37 | — | |
| STILTraining=10K Kinetics-400 + 10K Pika2026.05 | 72.5 | 45 | 62.07 | 98.88 | |
| STILTraining Data=Kinetics-400 and Pika, Evaluation Protocol=Standard2025.10 | 72.5 | 45 | 62.07 | 98.88 | |
| Skyra-SFTModel Size=7B2025.12 | 66.99 | 34.64 | 51.21 | — | |
| DeMamba2025.12 | 62.75 | 97.14 | 72.65 | — | |
| NSG-VD2025.12 | 50.22 | 99.28 | 66.6 | — | |
| AIGVDet2025.12 | 50 | 0 | 0 | — |