Text-to-Video Red-teaming on 390 Meta Harmful Seed Prompts
0Violence CountNaive
Evaluation results
| Method | Links | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| NaiveModel=Hunyuan-Video2025.11 | 0 | 1 | 2 | 0 | 4 | 3 | 10 | 98.3 | 98.9 | 99.4 | |
| NaiveModel=Wan 2.22025.11 | 2 | 1 | 1 | 0 | 3 | 2 | 9 | 98.6 | 97.2 | 98.3 | |
| UVDModel=Hunyuan-Video2025.11 | 25 | 21 | 20 | 8 | 17 | 22 | 113 | 90.8 | 90.2 | 91.1 | |
| UVDModel=Wan 2.22025.11 | 27 | 23 | 19 | 10 | 20 | 22 | 121 | 89 | 90 | 90.8 | |
| T2VSafetyBenchModel=Wan 2.22025.11 | 35 | 29 | 23 | 11 | 22 | 25 | 145 | 50.8 | 53.5 | 53.7 | |
| T2VSafetyBenchModel=Hunyuan-Video2025.11 | 37 | 30 | 24 | 18 | 21 | 29 | 159 | 51 | 53 | 54 | |
| ARTModel=Wan 2.22025.11 | 42 | 39 | 33 | 17 | 30 | 33 | 194 | 91.8 | 92.4 | 93.2 | |
| ARTModel=Hunyuan-Video2025.11 | 43 | 41 | 32 | 19 | 34 | 36 | 205 | 92.2 | 93 | 92.7 | |
| FLIRTModel=Wan 2.22025.11 | 45 | 44 | 39 | 20 | 36 | 36 | 220 | 48.6 | 50.8 | 50.5 | |
| FLIRTModel=Hunyuan-Video2025.11 | 47 | 44 | 37 | 21 | 36 | 38 | 223 | 51.4 | 52.2 | 51.9 | |
| TEARModel=Wan 2.22025.11 | 57 | 63 | 56 | 27 | 53 | 58 | 314 | 94.3 | 97.4 | 94.9 | |
| TEARModel=Hunyuan-Video2025.11 | 61 | 55 | 58 | 29 | 57 | 61 | 321 | 96.7 | 96.5 | 95.6 |