Prompt Injection Defense on AgentDojo Indirect Injection (test)
90Utility (No Attack)
Sandwich
Updated 1mo ago
Evaluation Results
| Method | Links | |||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
2026.04 | 90 | 80.89 | 1.95 | 77.32 | 1.33 | 75.36 | 1.28 | 79.5 | 1.7 | 74.05 | 1.11 | 79.25 | 1.95 | 58.15 | 17.61 | |
2026.04 | 85 | 76.18 | 2.78 | 70.15 | 2.61 | 71.71 | 2.45 | 68.14 | 2.13 | 73.31 | 2.91 | 75.73 | 3.57 | 57.74 | 45.14 | |
2026.04 | 85 | 65.15 | 1.69 | 74.85 | 0 | 75.32 | 0 | 74.84 | 0.11 | 76.73 | 0 | 78.25 | 0.11 | 63.75 | 0.69 | |
2026.04 | 82.5 | 82.65 | 2.22 | 75.12 | 1.06 | 77.12 | 1.76 | 78.13 | 1.11 | 79.49 | 1.59 | 76.31 | 2.23 | 58.22 | 46.79 | |
2026.04 | 82.5 | 56.69 | 2 | 65.59 | 0 | 69.19 | 0.25 | 77.9 | 0 | 69.94 | 0 | 70.31 | 0.22 | 35.42 | 1.04 | |
2026.04 | 80 | 71.83 | 2.72 | 76.34 | 0.56 | 72.3 | 0.54 | 73.29 | 0.49 | 75.75 | 0.43 | 78.32 | 1.67 | 60.66 | 44.77 | |
2026.04 | 77.5 | 63.65 | 1.95 | 77.88 | 0.28 | 77.03 | 1.84 | 79 | 0 | 77.93 | 0 | 71.56 | 1.11 | 68.75 | 22.15 | |
2026.04 | 62.5 | 57.66 | 1.33 | 63.87 | 0.78 | 63.98 | 1.15 | 68.98 | 0.15 | 65.25 | 0.78 | 60.16 | 0.4 | 61.02 | 0.28 | |
2026.04 | 40 | 12.5 | 0 | 18.4 | 0 | 11.49 | 0 | 10 | 0 | 12.56 | 0 | 25.18 | 0.27 | 19.75 | 3.55 |