Safety guardrailing on SafeEditBench
61.7Macro F1PolicyShiftGuard-7B
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| PolicyShiftGuard-7BParameter Scale=7B2026.07 | 61.7 | 69.9 | |
| PolicyShiftGuard-3BParameter Scale=3B2026.07 | 55.1 | 62.4 | |
| Qwen2.5-VL-7BParameter Scale=7B2026.07 | 48.7 | 32.9 | |
| SafeGuard-VL-RL-7BParameter Scale=7B2026.07 | 45.6 | 52.5 | |
| GuardReasoner-VL-3BParameter Scale=3B2026.07 | 41 | 51.8 | |
| GuardReasoner-VL-7BParameter Scale=7B2026.07 | 38.3 | 48.2 | |
| QwenGuard-7BParameter Scale=7B2026.07 | 32.8 | 38.7 | |
| LlamaGuard4-12BParameter Scale=12B2026.07 | 30.1 | 19.1 | |
| ShieldGemma2-4BParameter Scale=4B2026.07 | 27.5 | 36.8 |