Jailbreaking on HarmBench Visual Riddle
6.3ASR (K=1)Gemini 3.1 Pro
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Gemini 3.1 Pro2026.05 | 6.3 | 6.3 | |
| GPT-5.22026.05 | 13.2 | 13.2 | |
| Claude Haiku 4.52026.05 | 13.8 | 13.8 | |
| Qwen3-VL-235B2026.05 | 29.6 | 29.6 | |
| Qwen3-VL-32B2026.05 | 38.4 | 38.4 | |
| Gemini 3 Flash2026.05 | 52.2 | 52.2 |