Attack Success Rate Evaluation on HRL/LRL Safety Prompts Welsh Single Image v1
6ASRClaude 3 Haiku
Evaluation Results
| Method | Links | |
|---|---|---|
| Claude 3 Haiku2025.04 | 6 | |
| GPT-4o Mini2025.04 | 8 | |
| Gemini 1.5 Pro2025.04 | 8 | |
| Gemini 1.5 Flash2025.04 | 10 | |
| Claude 3.5 Sonnet2025.04 | 18 | |
| GPT-4o2025.04 | 86 |