LLM Agent Task Completion on AgentDojo No Attack
73.91Benign UtilityIPIGuard
Evaluation Results
| Method | Links | |
|---|---|---|
| IPIGuardModel=Gemini-2.52026.03 | 73.91 | |
| AttriGuardModel=Gemini-2.52026.03 | 67.39 | |
| SpotlightingModel=Llama3.3-70B2026.03 | 66.3 | |
| SandwichingModel=Gemini-2.52026.03 | 61.96 | |
| IPIGuardModel=Llama3.3-70B2026.03 | 57.61 | |
| SpotlightingModel=Gemini-2.52026.03 | 56.52 | |
| PromptArmorModel=Llama3.3-70B2026.03 | 56.52 | |
| SandwichingModel=Llama3.3-70B2026.03 | 56.52 | |
| PromptArmorModel=Gemini-2.52026.03 | 54.35 | |
| MELONModel=Gemini-2.52026.03 | 54.35 | |
| MELONModel=Llama3.3-70B2026.03 | 54.35 | |
| AttriGuardModel=Llama3.3-70B2026.03 | 54.35 | |
| DataSentinelModel=Llama3.3-70B2026.03 | 53.26 | |
| DataSentinelModel=Gemini-2.52026.03 | 51.09 | |
| PromptGuardModel=Llama3.3-70B2026.03 | 45.65 | |
| PromptGuardModel=Gemini-2.52026.03 | 40.22 | |
| CaMeLModel=Gemini-2.52026.03 | 38.04 | |
| PI DetectorModel=Llama3.3-70B2026.03 | 38.04 | |
| PIGuardModel=Llama3.3-70B2026.03 | 36.96 | |
| CaMeLModel=Llama3.3-70B2026.03 | 36.96 | |
| PIGuardModel=Gemini-2.52026.03 | 34.78 | |
| PI DetectorModel=Gemini-2.52026.03 | 32.61 |