Guarded Agent Evaluation on AgentHarm latest (full)
97.16Refusal Rate
ReAct-llamafirewall
Updated 1mo ago
Evaluation Results
| Method | Links | ||
|---|---|---|---|
2026.01 | 97.16 | 6.6 | |
2026.01 | 96.59 | 4.28 | |
2026.01 | 95.45 | 6.83 | |
2026.01 | 94.88 | 6.31 | |
2026.01 | 94.32 | 6.03 | |
2026.01 | 92.1 | 7.12 | |
2026.01 | 80.87 | 12.64 | |
2026.01 | 77.84 | 13.74 | |
2026.01 | 73.86 | 18.91 | |
2026.01 | 70.45 | 22.99 | |
2026.01 | 67.04 | 19.56 | |
2026.01 | 62.5 | 23.53 | |
2026.01 | 50 | 33.4 | |
2026.01 | 42.04 | 34.14 |