Multimodal Content Moderation on UnsafeBench Sexual Text+Visual
81.08AccuracyKidsNanny
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| KidsNannyRegime=2, Stage=1+22026.03 | 81.08 | 85.11 | 80.46 | 90.32 | 115 | |
| LlavaGuardRegime=22026.03 | 76.45 | 79.18 | 84.06 | 74.84 | 4,138 | |
| ShieldGemma-2Regime=22026.03 | 54.86 | 51.67 | 72.94 | 40 | 1,136 |