| Dataset Name | SOTA Method | Metric | Trend | ||
|---|---|---|---|---|---|
| MM-SafetyBench | CASA | Attack Success Rate (ASR)0.2 | 56 | 3mo ago | |
| FORTRESS | ReasoningGuard | ASR9.8 | 24 | 4mo ago | |
| PAIR | Self-Reminder | ASR1 | 24 | 2mo ago | |
| GCG | STAR-1 | ASR0 | 24 | 2mo ago | |
| Adversarial Jailbreak Prompts | Specreason | ATTEN Score63 | 20 | 1mo ago | |
| Mousetrap | Paraphrase | FFR (Reasoning)22.7 | 17 | 2mo ago | |
| AutoRAN | ThinkingI | Reasoning Failure Rate (FFR)0 | 17 | 2mo ago | |
| SorryBench | Paraphrase | FFR (Reasoning)56.8 | 17 | 2mo ago | |
| FigStep, MM-SafetyBench, SPA-VL Average | VSFA | Attack Success Rate (ASR)14.18 | 16 | 4mo ago | |
| SPA-VL | VSFA | Attack Success Rate (ASR)22.64 | 16 | 4mo ago | |
| ReNeLLM languageMix | Intention Analysis | ASR34 | 7 | 18d ago | |
| JailbreakV-28K v1 (test) | Defense Success Rate (Noise - T)38.16 | 6 | 4mo ago |