| Task Name | Dataset Name | SOTA Result | Trend | |
|---|---|---|---|---|
| Jailbreak Robustness | Jailbreak Attacks | Prefill Success Rate88.8 | 18 | |
| Jailbreak Defense | Jailbreak Attacks | GCG ASR0 | 18 | |
| Jailbreak Defense | Jailbreak Attacks | ES ASR28.1 | 10 | |
| Safety Defense Evaluation | Jailbreak Attacks GCG, AutoDAN, PAIR, DeepInc., SAP30, SIJ | GCG Attack Score1 | 8 | |
| Jailbreak Defense | Jailbreak Attacks ReNeLLM, DeepInception, GPTFuzzer, CodeAttack (test) | Harmful Score (ReNeLLM)1 | 6 | |
| Attack Detection | Jailbreak Attacks 105K sample set | Detection Rate68 | 4 |