Self-Harm on Generic Synthetic
2Bypass RateShieldgemma-9b
Evaluation Results
| Method | Links | |
|---|---|---|
| Shieldgemma-9bSafetyguard model=Shieldgemma-9b2026.05 | 2 | |
| Claude 4.5 Haiku2026.05 | 2 | |
| Gemini 2.5 flash2026.05 | 2 | |
| GPT o4-mini2026.05 | 2.25 | |
| Qwen3Guard-Gen-8BSafetyguard model=Qwen3Guard-Gen-8B2026.05 | 11.25 | |
| Llama 4 Scout2026.05 | 14 | |
| Llama-Guard-3-8BSafetyguard model=Llama-Guard-3-8B2026.05 | 35.63 |