GEO Attack Detection on ESCI Balanced (1:1) (test)
0FDRPrompt-Guard-2-86M
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Prompt-Guard-2-86MAttack=IOA2026.05 | 0 | 95.6 | |
| Prompt-Guard-2-86MAttack=Review2026.05 | 0 | 0.5 | |
| Prompt-Guard-2-86MAttack=Reason2026.05 | 0 | 0 | |
| Prompt-Guard-2-86MAttack=TAP2026.05 | 0 | 0 | |
| Prompt-Guard-2-86MAttack=SRP2026.05 | 0 | 5.8 | |
| Prompt-Guard-2-86MAttack=RAF2026.05 | 0 | 1 | |
| Prompt-Guard-2-86MAttack=STS2026.05 | 0 | 80.6 | |
| PG-FTAttack=IOA2026.05 | 2.7 | 98.6 | |
| PG-FTAttack=Review2026.05 | 2.7 | 98.1 | |
| PG-FTAttack=Reason2026.05 | 2.7 | 98.6 | |
| PG-FTAttack=SRP2026.05 | 2.7 | 98.4 | |
| PG-FTAttack=STS2026.05 | 2.7 | 98.6 | |
| PG-FTAttack=TAP2026.05 | 2.8 | 95.4 | |
| PG-FTAttack=RAF2026.05 | 3.1 | 90.4 | |
| Qwen3Guard-Gen-8BAttack=TAP2026.05 | 3.5 | 42.3 | |
| Llama-Guard-4-12BAttack=STS2026.05 | 11.8 | 73.8 | |
| Llama-Guard-4-12BAttack=SRP2026.05 | 34.6 | 25.7 | |
| Llama-Guard-4-12BAttack=TAP2026.05 | 37.7 | 19.7 | |
| Llama-Guard-4-12BAttack=RAF2026.05 | 39.4 | 21.4 | |
| Llama-Guard-4-12BAttack=IOA2026.05 | 39.9 | 21 | |
| Qwen3Guard-Gen-8BAttack=IOA2026.05 | 50 | 1.5 | |
| Qwen3Guard-Gen-8BAttack=STS2026.05 | 50 | 1.5 | |
| Llama-Guard-4-12BAttack=Review2026.05 | 56.5 | 11.3 | |
| Llama-Guard-4-12BAttack=Reason2026.05 | 56.5 | 11.3 | |
| Qwen3Guard-Gen-8BAttack=Review2026.05 | 75 | 1 | |
| Qwen3Guard-Gen-8BAttack=Reason2026.05 | 75 | 1 | |
| Qwen3Guard-Gen-8BAttack=SRP2026.05 | 75 | 1 | |
| Qwen3Guard-Gen-8BAttack=RAF2026.05 | 75 | 1 |