Safety Evaluation Efficiency on Unsafe and Jailbreak Prompts
13.29RAS Time (s)Gemma-3-4b-it
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| Gemma-3-4b-itModel Family=Gemma2026.06 | 13.29 | 6,309.68 | 474.85 | |
| Llama-3.1-8B-InstructModel Family=Llama2026.06 | 14.13 | 572.47 | 40.52 | |
| AverageScope=Cross-family average2026.06 | 14.13 | 2,969.14 | 210.13 | |
| Qwen2.5-7B-InstructModel Family=Qwen2026.06 | 14.97 | 2,025.26 | 135.25 |