Tool-use LLM Agent Task Completion on AgentDojo Weighted Average (Overall)
85.57Benign Success RateRepeat User Prompt
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| Repeat User PromptDefense Status=With Defense, Backbone=GPT-4o2026.06 | 85.57 | 57.73 | 43.3 | |
| System PromptDefense Status=No Defense, Backbone=GPT-4o2026.06 | 80.41 | 44.33 | 42.27 | |
| Tool FilterDefense Status=No Defense, Backbone=GPT-4o2026.06 | 78.35 | 48.45 | 42.27 | |
| System PromptDefense Status=With Defense, Backbone=GPT-4o2026.06 | 76.29 | 64.95 | 19.59 | |
| Repeat User PromptDefense Status=No Defense, Backbone=GPT-4o2026.06 | 76.29 | 50.52 | 39.18 | |
| DriftDefense Status=No Defense, Backbone=GPT-4o2026.06 | 76.29 | 54.64 | 20.62 | |
| Tool-GuardDefense Status=No Defense, Backbone=GPT-4o2026.06 | 76.29 | 54.64 | 43.3 | |
| ProagentDefense Status=No Defense, Backbone=GPT-4o2026.06 | 73.2 | 48.45 | 31.96 | |
| Tool-GuardDefense Status=With Defense, Backbone=GPT-4o2026.06 | 72.16 | 58.76 | 2.06 | |
| Tool FilterDefense Status=With Defense, Backbone=GPT-4o2026.06 | 69.07 | 55.67 | 10.31 | |
| DriftDefense Status=With Defense, Backbone=GPT-4o2026.06 | 68.04 | 44.33 | 24.74 | |
| ProagentDefense Status=With Defense, Backbone=GPT-4o2026.06 | 43.3 | 13.4 | 19.59 |