OpenAI Moderation
Benchmarks
| Task Name | Dataset Name | SOTA Result | Trend | |
|---|---|---|---|---|
| Harmfulness Detection | OpenAI Moderation | Macro F1 Score92.9 | 45 | |
| Safety Classification | OpenAI-moderation (test) | Accuracy74.88 | 23 | |
| Prompt Classification | OpenAI Moderation Text Prompt | F1 Score88.89 | 14 | |
| Unsafe content categorization | OpenAI Moderation | Accuracy88.35 | 9 | |
| Multi-label Safety Categorization | OpenAI Moderation | Macro Accuracy47.67 | 4 | |
| Out-of-Taxonomy Risk Detection | OpenAI Moderation | Out-of-Taxonomy F167.92 | 4 | |
| OOD safety category inference (Stage 2) | OpenAI Moderation | Mean Reward36.45 | 4 | |
| Jailbreaking | OpenAI’s Moderation | Bypass Rate100 | 1 |
Showing 8 of 8 rows