List your company

Prompt Injection Defense on AgentDojo New Attack 1

89.88Utility under AttackNo Defense
5.17227.163549.15571.1465Feb 8, 2026

Evaluation results

MethodLinks
No DefenseBackbone Model=Gemini-3-Pro2026.0289.8815.69
Repeat PromptBackbone Model=Gemini-3-Pro2026.0289.441.174.11
CausalArmorBackbone Model=Gemini-3-Pro2026.0288.091.430.11
No DefenseBackbone Model=Claude-4-Sonnet2026.0285.7713.37
CausalArmorBackbone Model=Claude-4-Sonnet2026.0283.671.560.11
PromptArmorBackbone Model=Gemini-3-Pro2026.0283.672.010.11
PromptArmorBackbone Model=Claude-4-Sonnet2026.0274.292.40
DRIFTBackbone Model=Gemini-3-Pro2026.0271.456.613.65
Repeat PromptBackbone Model=Claude-4-Sonnet2026.0268.61.211.26
DRIFTBackbone Model=Claude-4-Sonnet2026.0265.316.380.63
DeBERTa DetectorBackbone Model=Claude-4-Sonnet2026.0262.81.111.79
DeBERTa DetectorBackbone Model=Gemini-3-Pro2026.0256.311.053.37
CausalArmorBackbone Model=Gemini-2.5-Flash2026.0254.641.780.42
No DefenseBackbone Model=Gemini-2.5-Flash2026.0251.59112.01
PromptArmorBackbone Model=Gemini-2.5-Flash2026.0250.322.50
Repeat PromptBackbone Model=Gemini-2.5-Flash2026.0250.011.158.51
MelonBackbone Model=Claude-4-Sonnet2026.0241.842.480.21
DRIFTBackbone Model=Gemini-2.5-Flash2026.0236.385.582.48
DeBERTa DetectorBackbone Model=Gemini-2.5-Flash2026.0233.621.24.21
MelonBackbone Model=Gemini-2.5-Flash2026.0228.723.120.21
PiGuardBackbone Model=Gemini-3-Pro2026.0225.921.060.63
PiGuardBackbone Model=Claude-4-Sonnet2026.0225.41.150
PiGuardBackbone Model=Gemini-2.5-Flash2026.028.431.290.42