Safety Evaluation on MCV SafetyBench
34.381-Clip Safety ScoreGemini-2.5-Flash
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| Gemini-2.5-FlashAttack=Explicit2026.06 | 34.38 | 35.07 | 35.75 | 35.07 | 35.89 | |
| Gemini-2.5-FlashAttack=Implicit2026.06 | 34.32 | 35.14 | 37.12 | 40.48 | 41.3 | |
| GPT-5-miniAttack=Implicit2026.06 | 26.23 | 25.62 | 26.51 | 26.44 | 23.29 | |
| GPT-5-miniAttack=Explicit2026.06 | 23.15 | 22.67 | 24.25 | 24.79 | 24.86 |