AI-generated video detection on DeepTraceReward
97.2AccuracyQwen2.5-ViT
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| Qwen2.5-ViT2026.04 | 97.2 | 96.3 | 98.2 | |
| GPT-4.12026.04 | 92.9 | 89.1 | 97.9 | |
| GPT-52026.04 | 90.7 | 84.6 | 98.8 | |
| Gemini 2.5 Pro2026.04 | 84.3 | 75.7 | 95.8 | |
| DeepTraceRewardBackbone=Qwen2.5 VL 7B2026.04 | 74.7 | 55.7 | 100 | |
| Qwen2.5-VLModel Size=7B2026.04 | 51.7 | 20.2 | 93.4 | |
| Qwen2.5-VLModel Size=72B2026.04 | 50 | 16.6 | 94.3 | |
| Qwen2.5-VLModel Size=32B2026.04 | 47.4 | 8.9 | 98.5 | |
| VideoLLaMa3Model Size=7B2026.04 | 10 | 38.1 | 4.3 |