Malicious Instruction Detection on MaliciousAgentSkillsBench (traditional IPI baselines)
63.93PrecisionRouteGuard
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| RouteGuard2026.04 | 63.93 | 57.5 | 78.67 | |
| RENNERVATE2026.04 | 47.76 | 55.28 | 70.17 | |
| ASen2026.04 | 36.76 | 22.5 | 27.17 | |
| PArm2026.04 | 24 | 20 | 21.2 |