Prompt Injection Defense on CSQA random architecture
18.3ASRPropGuard
Evaluation results
| Method | Links | ||
|---|---|---|---|
| PropGuardBase Model=GPT-4o-mini, Communication Architecture=random2026.05 | 18.3 | 83.3 | |
| Qwen3GuardBase Model=GPT-4o-mini, Communication Architecture=random2026.05 | 25.3 | 78.3 | |
| ThinkGuardBase Model=GPT-4o-mini, Communication Architecture=random2026.05 | 27.3 | 71.7 | |
| WildGuardBase Model=GPT-4o-mini, Communication Architecture=random2026.05 | 30.7 | 65 | |
| LlamaGuardBase Model=GPT-4o-mini, Communication Architecture=random2026.05 | 36.3 | 61.7 | |
| No DefenseBase Model=GPT-4o-mini, Communication Architecture=random2026.05 | 44.7 | 55 |