| Dataset Name | SOTA Method | Metric | Trend | ||
|---|---|---|---|---|---|
| MM-Safety Bench (test) | LLaVA-v1.5-7B | Average ASR0.18 | 56 | 3mo ago | |
| JailbreakBench | Attack Success Rate (J)2 | 9 | 1mo ago | ||
| English dataset Multi-Image | StrongREJECT (Perturbed)14 | 6 | 3mo ago | ||
| English dataset Single-Image | StrongREJECT (Perturbed)10 | 6 | 3mo ago | ||
| English dataset Text | StrongREJECT Rate0.01 | 6 | 3mo ago | ||
| SHAPE | Gemini 2.5 Flash-Lite | Cipher Success Rate100 | 5 | 1mo ago | |
| StrongREJECT (test) | STAIR-DPO | Overall Score100 | 5 | 2mo ago |