Jailbreaking on HarmBench Textual Riddle
67.9ASR (K=1)Gemini 3 Flash
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Gemini 3 Flash2026.05 | 67.9 | 67.9 | |
| Qwen3-VL-32B2026.05 | 62.3 | 62.3 | |
| Qwen3-VL-235B2026.05 | 51.6 | 51.6 | |
| Claude Haiku 4.52026.05 | 39.6 | 39.6 | |
| GPT-5.22026.05 | 24.5 | 24.5 | |
| Gemini 3.1 Pro2026.05 | 17 | 17 |