Safety Classification on WildGuard overall (test)
85.6AccuracyReflect-Guard
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| Reflect-GuardModel=Reflect-Guard (ours)2026.05 | 85.6 | 82.1 | 86.3 | 84.2 | 85.5 | |
| Llama-Guard-3-8BModel=Llama-Guard-3-8B (baseline)2026.05 | 82.5 | 91.9 | 66.3 | 77 | 70.2 |