| Dataset Name | SOTA Method | Metric | Trend | ||
|---|---|---|---|---|---|
| StrongREJECT | PPO SafeSearch | Harm Rate0.2 | 29 | 2mo ago | |
| VLSBench | SAFER-STEER | Safety Score84.41 | 18 | 1mo ago | |
| VLGuard | SAFER-STEER | Safety Score88.48 | 18 | 1mo ago | |
| SPA-VL | SAFER-STEER | Safety Score85.8 | 18 | 1mo ago | |
| MM-Safety | SAFER-STEER | Safety Score88.71 | 18 | 1mo ago | |
| Beavertails | SAFER-STEER | Safety Score92.03 | 18 | 1mo ago | |
| WildTeaming | PPO SafeSearch w/o qr. & hf. | Harm Rate0.3 | 15 | 2mo ago | |
| RRB | GRPO SafeSearch | HarmR1 | 15 | 2mo ago | |
| XSTest | XSTest Score74.8 | 11 | 3mo ago | ||
| Phone-Harm Harm-150 | GPT-5 | HR0.4 | 8 | 1mo ago |