List your company

Prompt Injection Defense on AgentDojo Important Instructions

0.9041Utility under AttackNo Defense
0.047140.269620.49210.71458Feb 8, 2026

Evaluation results

MethodLinks
No DefenseBackbone Model=Gemini-3-Pro2026.020.904110.0506
Repeat PromptBackbone Model=Gemini-3-Pro2026.020.89991.180.0285
CausalArmorBackbone Model=Gemini-3-Pro2026.020.87141.440.0011
No DefenseBackbone Model=Claude-4-Sonnet2026.020.839810.0232
CausalArmorBackbone Model=Claude-4-Sonnet2026.020.83671.540
PromptArmorBackbone Model=Gemini-3-Pro2026.020.83351.980.0011
PromptArmorBackbone Model=Claude-4-Sonnet2026.020.74082.440
DRIFTBackbone Model=Gemini-3-Pro2026.020.70656.440.0385
Repeat PromptBackbone Model=Claude-4-Sonnet2026.020.69021.180.0084
DRIFTBackbone Model=Claude-4-Sonnet2026.020.66926.630.0082
DeBERTa DetectorBackbone Model=Gemini-3-Pro2026.020.55432.060.0105
CausalArmorBackbone Model=Gemini-2.5-Flash2026.020.5091.960.0062
DeBERTa DetectorBackbone Model=Claude-4-Sonnet2026.020.49631.120.0137
PromptArmorBackbone Model=Gemini-2.5-Flash2026.020.41822.650.0063
Repeat PromptBackbone Model=Gemini-2.5-Flash2026.020.41411.150.3593
MelonBackbone Model=Claude-4-Sonnet2026.020.41412.410.0021
No DefenseBackbone Model=Gemini-2.5-Flash2026.020.324610.3247
PiGuardBackbone Model=Claude-4-Sonnet2026.020.3131.160.0063
PiGuardBackbone Model=Gemini-3-Pro2026.020.31191.060.0063
DRIFTBackbone Model=Gemini-2.5-Flash2026.020.16876.750.0499
DeBERTa DetectorBackbone Model=Gemini-2.5-Flash2026.020.16751.240.0727
MelonBackbone Model=Gemini-2.5-Flash2026.020.1382.830.0053
PiGuardBackbone Model=Gemini-2.5-Flash2026.020.08011.280.0105