AI-generated video detection on GenVideo Lavie
94.32ACCSkyra-RL-GenVideo
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| Skyra-RL-GenVideoModel Size=7B2025.12 | 94.32 | 0.9429 | 0.9432 | — | |
| NSG-VDTraining=10K Kinetics-400 + 10K Pika2026.05 | 94.17 | 0.9417 | 0.9417 | 97.41 | |
| NSG-VDTraining Data=Kinetics-400 and Pika, Evaluation Protocol=Standard2025.10 | 94.17 | 0.9417 | 0.9417 | 97.41 | |
| CAM-VFDTraining=10K Kinetics-400 + 10K Pika2026.05 | 93.6 | 0.884 | 0.91 | 96.9 | |
| NSG-VDTraining Data=Kinetics-400 and SEINE2025.10 | 87.92 | 0.975 | 0.8897 | 95.09 | |
| DeMambaTraining Data=Kinetics-400 and SEINE2025.10 | 83.4 | 0.686 | 0.8052 | 96.87 | |
| Skyra-RLModel Size=7B2025.12 | 78.5 | 0.5914 | 0.7334 | — | |
| DeMambaTraining=10K Kinetics-400 + 10K Pika2026.05 | 77.7 | 0.59 | 0.7257 | 91.32 | |
| DeMambaTraining Data=Kinetics-400 and Pika, Evaluation Protocol=Standard2025.10 | 77.7 | 0.59 | 0.7257 | 91.32 | |
| NPRTraining Data=Kinetics-400 and SEINE2025.10 | 77.6 | 0.588 | 0.7241 | 89.46 | |
| TALLTraining=10K Kinetics-400 + 10K Pika2026.05 | 74.1 | 0.492 | 0.6551 | 94.84 | |
| TALLTraining Data=Kinetics-400 and Pika, Evaluation Protocol=Standard2025.10 | 74.1 | 0.492 | 0.6551 | 94.84 | |
| TALLTraining Data=Kinetics-400 and SEINE2025.10 | 71.4 | 0.438 | 0.605 | 94.8 | |
| Skyra-SFTModel Size=7B2025.12 | 67.86 | 0.3629 | 0.5303 | — | |
| NPRTraining=10K Kinetics-400 + 10K Pika2026.05 | 66.5 | 0.346 | 0.5081 | 87.54 | |
| NPRTraining Data=Kinetics-400 and Pika, Evaluation Protocol=Standard2025.10 | 66.5 | 0.346 | 0.5081 | 87.54 | |
| STILTraining Data=Kinetics-400 and SEINE2025.10 | 62.3 | 0.248 | 0.3968 | 91 | |
| DeMamba2025.12 | 61.29 | 0.9886 | 0.7167 | — | |
| STILTraining=10K Kinetics-400 + 10K Pika2026.05 | 53.6 | 0.072 | 0.1343 | 88.16 | |
| STILTraining Data=Kinetics-400 and Pika, Evaluation Protocol=Standard2025.10 | 53.6 | 0.072 | 0.1343 | 88.16 | |
| NSG-VD2025.12 | 50.29 | 0.9921 | 0.6662 | — | |
| AIGVDet2025.12 | 50.04 | 0.0007 | 0.0014 | — |