Safety Jailbreak Evaluation on Forbidden Questions
53.6ASRBaseline (No Attack)
Evaluation Results
| Method | Links | |
|---|---|---|
| Baseline (No Attack)Model=Llama-2-7B-Chat, Setting=Baseline (No Attack)2026.03 | 53.6 | |
| Baseline (No Attack)Model=Llama-3-8B-Instruct, Setting=Baseline (No Attack)2026.03 | 69.2 | |
| AmnesiaModel=Llama-3-8B-Instruct, Setting=Amnesia (Layers 13,14,15)2026.03 | 92.3 |