Jailbreak Attack Evaluation on Paired Prompts Held-out (test)
0.78Similarity
PAIR
Updated 22d ago
Evaluation Results
| Method | Links | ||||||||
|---|---|---|---|---|---|---|---|---|---|
2026.05 | 0.78 | 43.56 | 0.24 | 0.3 | 0.44 | 0.08 | 0.72 | 0.64 | |
2026.05 | 0.53 | 42.23 | 0.23 | 0.32 | 0.43 | 0.09 | 0.84 | 0.66 | |
2026.05 | 0.48 | 20.54 | 0.24 | 0.3 | - | 0.16 | 1 | 1 | |
2026.05 | 0.4 | 33.37 | 0.25 | 0.33 | 0.41 | 0.15 | 1 | 1 | |
2026.05 | 0.35 | 16.03 | 0.25 | 0.32 | 0.42 | 0.14 | 0.96 | 0.98 | |
2026.05 | 0.2 | 957.51 | 0.25 | 0.31 | 0.46 | 0.05 | 1 | 0.96 | |
2026.05 | 0.1 | 95.55 | - | - | - | 0.02 | 1 | 1 | |
2026.05 | 0.08 | 34.39 | - | - | - | 0.09 | 1 | 1 |