LLM Agent Security Defense on AgentDojo Dynamic Benchmark
77.32UtilityNone/baseline
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| None/baselineModel=gpt-5-mini, Time (s)=4.032026.03 | 77.32 | 47.69 | |
| Agent-SentryModel=gpt-5-mini, Time (s)=7.892026.03 | 76.3 | 3.7 | |
| ProgentModel=gpt-5-mini, Time (s)=4.262026.03 | 73.2 | 5.41 | |
| CaMeLModel=gpt-5-mini, Time (s)=26.602026.03 | 37.11 | 0 |