Jailbreaking on AdvBench Baseline No System Prompt brief-variant
0Jailbreaks CountPhi-4
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Phi-4Open-weight=true, System Prompt=None2026.04 | 0 | 0 | |
| GPT-5-MiniOpen-weight=false, System Prompt=None2026.04 | 1 | 0.19 | |
| Claude-Haiku-4.5Open-weight=false, System Prompt=None2026.04 | 2 | 0.38 | |
| Gemini-3-FlashOpen-weight=false, System Prompt=None2026.04 | 4 | 0.77 | |
| Command-R7BOpen-weight=true, System Prompt=None2026.04 | 5 | 0.96 | |
| Grok-4.1Open-weight=false, System Prompt=None2026.04 | 8 | 1.54 | |
| Mistral-Medium-3.1Open-weight=false, System Prompt=None2026.04 | 8 | 1.54 | |
| Qwen3-8BOpen-weight=true, System Prompt=None2026.04 | 11 | 2.12 | |
| Gemma-3-12BOpen-weight=true, System Prompt=None2026.04 | 18 | 3.46 | |
| Llama-3.1-8BOpen-weight=true, System Prompt=None2026.04 | 37 | 7.12 |