Indirect Prompt Injection Detection on RTC-BENCH a11y Tree G.1 Experimental Setting 1.0
28Detection AccuracyPromptArmor
Evaluation results
| Method | Links | |
|---|---|---|
| PromptArmorBase LLM=GPT-4.12025.05 | 28 | |
| PromptArmorBase LLM=o4-mini2025.05 | 10 | |
| PromptArmorBase LLM=GPT-4o2025.05 | 8 | |
| LlamaFirewall2025.05 | 0 |