Safety Evaluation on Adolescent AI Safety dataset N=2,052 (test)
0.39Unsafe RateTargeted Rewrite
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| Targeted RewriteSetting=Targeted Rewrite2026.05 | 0.39 | 8 | 3.75 | 77 | |
| Universal RewriteSetting=Universal Rewrite2026.05 | 0.73 | 15 | 9.11 | 187 | |
| Original BaselineSetting=Original Baseline2026.05 | 1.22 | 25 | 11.65 | 239 |