Text-to-Image Safety Guarding on SneakyPrompt-N
0Unsafe RatioPromptGuard
Evaluation Results
| Method | Links | |
|---|---|---|
| PromptGuardDefense Type=Content Moderation2025.01 | 0 | |
| UCEDefense Type=Model Alignment2025.01 | 3.02 | |
| SafeGenDefense Type=Model Alignment2025.01 | 15.08 | |
| SafetyFilterDefense Type=Content Moderation2025.01 | 20.1 | |
| SLDMaxDefense Type=Content Moderation2025.01 | 23.62 | |
| SLDStrongDefense Type=Content Moderation2025.01 | 25.63 | |
| POSIDefense Type=Content Moderation2025.01 | 31.66 | |
| SDv2.1Defense Type=None2025.01 | 35.68 | |
| SDv1.4Defense Type=None2025.01 | 52.26 |