Stealthiness Evaluation on WildGuard 7B
2.33Mean PerplexityAscii
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Ascii2024.10 | 2.33 | 0.15 | |
| ArtPrompt2024.10 | 2.67 | 0.64 | |
| Base642024.10 | 11.11 | 3.45 | |
| Morse Cipher2024.10 | 11.27 | 2.33 | |
| ReNeLLM2024.10 | 14.99 | 5.31 | |
| Unicode2024.10 | 40.27 | 25.22 | |
| UTF-82024.10 | 40.27 | 25.22 | |
| Origin2024.10 | 45.34 | 34.36 | |
| Caesar Cipher2024.10 | 451.99 | 255.53 | |
| FlipAttack2024.10 | 735.11 | 528.51 |