Indirect Prompt Injection Defense on AgentDojo (Full)
78.26BU (NOATTACK)MetaSecAlign
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| MetaSecAlignModel=Llama3.3-70B2026.03 | 78.26 | 78.91 | 0.79 | 77.32 | 0.79 | |
| No DefenseModel=Llama3.3-70B2026.03 | 59.78 | 43.08 | 22.11 | 47.51 | 18.82 | |
| AttriGuardModel=Llama3.3-70B2026.03 | 54.25 | 47.62 | 0 | 45.8 | 0 |