Tool-use LLM Agent Task Completion on AgentDojo Travel suite
75Benign SuccessSystem Prompt
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| System PromptDefense Status=No Defense, Backbone=GPT-4o2026.06 | 75 | 10 | 90 | |
| System PromptDefense Status=With Defense, Backbone=GPT-4o2026.06 | 75 | 60 | 30 | |
| Tool FilterDefense Status=With Defense, Backbone=GPT-4o2026.06 | 75 | 70 | 10 | |
| Repeat User PromptDefense Status=No Defense, Backbone=GPT-4o2026.06 | 75 | 30 | 60 | |
| DriftDefense Status=No Defense, Backbone=GPT-4o2026.06 | 75 | 35 | 0 | |
| Tool FilterDefense Status=No Defense, Backbone=GPT-4o2026.06 | 70 | 15 | 90 | |
| Repeat User PromptDefense Status=With Defense, Backbone=GPT-4o2026.06 | 65 | 20 | 75 | |
| DriftDefense Status=With Defense, Backbone=GPT-4o2026.06 | 65 | 10 | 0 | |
| ProagentDefense Status=With Defense, Backbone=GPT-4o2026.06 | 65 | 10 | 75 | |
| Tool-GuardDefense Status=No Defense, Backbone=GPT-4o2026.06 | 65 | 20 | 85 | |
| Tool-GuardDefense Status=With Defense, Backbone=GPT-4o2026.06 | 65 | 55 | 0 | |
| ProagentDefense Status=No Defense, Backbone=GPT-4o2026.06 | 60 | 5 | 100 |