Safety Evaluation on HarmBench Static Context Evaluation Single-turn
70.5SFRCodeAttack
Evaluation Results
| Method | Links | |
|---|---|---|
| CodeAttackTarget Model=GPT 4o, Access Type=Closed-Source Models2026.03 | 70.5 | |
| CodeAttackTarget Model=Llama 3-70B-IT, Access Type=Open-Weight Models2026.03 | 66 | |
| CodeAttackTarget Model=Llama 3-8B-IT, Access Type=Open-Weight Models2026.03 | 46 | |
| CodeAttackTarget Model=Claude 3.5 Sonnet, Access Type=Closed-Source Models2026.03 | 39.5 | |
| PAIRTarget Model=GPT 4o, Access Type=Closed-Source Models2026.03 | 39 | |
| PAIRTarget Model=Llama 3-70B-IT, Access Type=Open-Weight Models2026.03 | 36 | |
| GCGTarget Model=Llama 3-8B-IT, Access Type=Open-Weight Models2026.03 | 34.5 | |
| PAIRTarget Model=Llama 3-8B-IT, Access Type=Open-Weight Models2026.03 | 18.7 | |
| GCGTarget Model=Llama 3-70B-IT, Access Type=Open-Weight Models2026.03 | 17 | |
| GCGTarget Model=GPT 4o, Access Type=Closed-Source Models2026.03 | 12.5 | |
| GCGTarget Model=Claude 3.5 Sonnet, Access Type=Closed-Source Models2026.03 | 3 | |
| PAIRTarget Model=Claude 3.5 Sonnet, Access Type=Closed-Source Models2026.03 | 3 |