AI-generated video detection on GenVideo Wild Scrape
92.3AccuracyCAM-VFD
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| CAM-VFDTraining=10K Kinetics-400 + 10K Pika2026.05 | 92.3 | 87.9 | 90.1 | 96.2 | |
| NSG-VDTraining=10K Kinetics-400 + 10K Pika2026.05 | 88.75 | 82.5 | 88 | 94.73 | |
| NSG-VDTraining Data=Kinetics-400 and Pika, Evaluation Protocol=Standard2025.10 | 88.75 | 82.5 | 88 | 94.73 | |
| Skyra-RL-GenVideoModel Size=7B2025.12 | 86.56 | 78.63 | 85.41 | — | |
| NSG-VDTraining Data=Kinetics-400 and SEINE2025.10 | 78.33 | 89.17 | 80.45 | 88.65 | |
| DeMambaTraining=10K Kinetics-400 + 10K Pika2026.05 | 77.3 | 58.2 | 71.94 | 87.38 | |
| DeMambaTraining Data=Kinetics-400 and Pika, Evaluation Protocol=Standard2025.10 | 77.3 | 58.2 | 71.94 | 87.38 | |
| TALLTraining=10K Kinetics-400 + 10K Pika2026.05 | 76.3 | 53.6 | 69.34 | 93.75 | |
| TALLTraining Data=Kinetics-400 and Pika, Evaluation Protocol=Standard2025.10 | 76.3 | 53.6 | 69.34 | 93.75 | |
| DeMambaTraining Data=Kinetics-400 and SEINE2025.10 | 73.1 | 48 | 64.09 | 88.11 | |
| NPRTraining=10K Kinetics-400 + 10K Pika2026.05 | 70.8 | 43.2 | 59.67 | 91.84 | |
| NPRTraining Data=Kinetics-400 and Pika, Evaluation Protocol=Standard2025.10 | 70.8 | 43.2 | 59.67 | 91.84 | |
| Skyra-RLModel Size=7B2025.12 | 68.34 | 38.88 | 55.11 | — | |
| TALLTraining Data=Kinetics-400 and SEINE2025.10 | 67.4 | 35.8 | 52.34 | 89.45 | |
| NPRTraining Data=Kinetics-400 and SEINE2025.10 | 61.9 | 27.4 | 41.83 | 76.66 | |
| Skyra-SFTModel Size=7B2025.12 | 61.62 | 24.56 | 39.02 | — | |
| STILTraining Data=Kinetics-400 and SEINE2025.10 | 59.4 | 19 | 31.88 | 86.58 | |
| STILTraining=10K Kinetics-400 + 10K Pika2026.05 | 55.8 | 11.6 | 20.79 | 87.52 | |
| STILTraining Data=Kinetics-400 and Pika, Evaluation Protocol=Standard2025.10 | 55.8 | 11.6 | 20.79 | 87.52 | |
| DeMamba2025.12 | 55.51 | 88.11 | 66.45 | — | |
| AIGVDet2025.12 | 50 | 0 | 0 | — | |
| NSG-VD2025.12 | 40.94 | 99.36 | 57.86 | — |