Prompt Injection Defense on MATH random architecture
6ASRPropGuard
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| PropGuardBase Model=GPT-4o-mini, Communication Architecture=random2026.05 | 6 | 95 | |
| Qwen3GuardBase Model=GPT-4o-mini, Communication Architecture=random2026.05 | 15.3 | 86.7 | |
| ThinkGuardBase Model=GPT-4o-mini, Communication Architecture=random2026.05 | 16.7 | 83.3 | |
| WildGuardBase Model=GPT-4o-mini, Communication Architecture=random2026.05 | 22.3 | 78.3 | |
| LlamaGuardBase Model=GPT-4o-mini, Communication Architecture=random2026.05 | 23.7 | 78.3 | |
| No DefenseBase Model=GPT-4o-mini, Communication Architecture=random2026.05 | 26.7 | 71.7 |