False Alarm Detection on AlpacaEval
0False Alarm Rate (FAR)Llama-Guard-3
Evaluation Results
| Method | Links | |
|---|---|---|
| Llama-Guard-3Target Model=Vicuna2026.05 | 0 | |
| Llama-Guard-3Target Model=Llama22026.05 | 0 | |
| Llama-Guard-3Target Model=Qwen2.52026.05 | 0 | |
| embedding disruption methodTarget Model=Qwen2.52026.05 | 0 | |
| embedding disruption methodTarget Model=Vicuna2026.05 | 1 | |
| embedding disruption methodTarget Model=Llama22026.05 | 1 | |
| GuardReasonerTarget Model=Vicuna2026.05 | 38 | |
| GuardReasonerTarget Model=Llama22026.05 | 38 | |
| GuardReasonerTarget Model=Qwen2.52026.05 | 38 |