| Dataset Name | SOTA Method | Metric | Trend | ||
|---|---|---|---|---|---|
| PolicyGuardBench | Safety F188.83 | 30 | 2mo ago | ||
| WildGuard (test) | GuardReasoner-8B | Safety F189.17 | 7 | 2mo ago | |
| HarmBench | LPG-4B (Ours) | Safety F196.44 | 7 | 2mo ago | |
| StepShield (main) | EIR18 | 4 | 5mo ago | ||
| 73 real DPA enforcement decisions V4 | Rule B | Fires28 | 2 | 1mo ago |