| Task Name | Dataset Name | SOTA Result | Trend | |
|---|---|---|---|---|
| Jailbreak Defense | GCG | Harmful Score1 | 37 | |
| Jailbreak Detection | GCG | Accuracy99 | 30 | |
| Jailbreak Attack | GCG | ASR96 | 27 | |
| Jailbreak Attack Defense | GCG | ASR0 | 24 | |
| Adversarial Robustness | GCG | GCG Rate0.13 | 21 | |
| Adversarial Attack Defense | GCG Individual | BAR100 | 18 | |
| Interleaved text-mask generation | GCG (test) | METEOR17.4 | 10 | |
| Interleaved text-mask generation | GCG (val) | METEOR17.7 | 10 | |
| Prompt Injection | GCG Clean | ASR37.02 | 4 | |
| Grounded Conversation Generation | GCG (test) | mIoU62.34 | 3 |