Unsafe content categorization on Aegis 2.0
63.18AccuracyGemini2.5-Flash
Evaluation Results
| Method | Links | |
|---|---|---|
| Gemini2.5-Flash2025.12 | 63.18 | |
| GPT4o-mini2025.12 | 57.58 | |
| ProGuard-7B2025.12 | 49.04 | |
| ProGuard-3B2025.12 | 35.63 | |
| LlamaGuard2-8B2025.12 | 26.45 | |
| LlamaGuard4-12B2025.12 | 25.25 | |
| LlamaGuard3-8B2025.12 | 24.79 | |
| LlamaGuard3-11B-Vision2025.12 | 17.91 | |
| LlamaGuard-7B2025.12 | 14.42 |