Human-crafted jailbreak defense on mhj
14Attack Success Rate (ASR)
Cognitive Firewall
Updated 22d ago
Evaluation Results
| Method | Links | ||
|---|---|---|---|
2026.07 | 14 | 86 | |
2026.07 | 26 | 66 | |
2026.07 | 38 | 54 | |
2026.07 | 40 | 50 | |
2026.07 | 44 | 40 | |
2026.07 | 50 | 40 | |
2026.07 | 74 | 0 |
Cognitive Firewall
| Method | Links | ||
|---|---|---|---|
2026.07 | 14 | 86 | |
2026.07 | 26 | 66 | |
2026.07 | 38 | 54 | |
2026.07 | 40 | 50 | |
2026.07 | 44 | 40 | |
2026.07 | 50 | 40 | |
2026.07 | 74 | 0 |