| Task Name | Dataset Name | SOTA Result | Trend | |
|---|---|---|---|---|
| Safety Evaluation | Llama-Guard 3-8B | ASR2.38 | 56 | |
| Safety alignment evaluation | Llama-Guard | Harmfulness (%)82.14 | 36 | |
| Stealthiness Evaluation | LLaMA Guard 8B 3.1 | Mean PPL2.16 | 10 | |
| Stealthiness Evaluation | LLaMA Guard 2 8B | PPL Mean2.34 | 10 | |
| Stealthiness Evaluation | LLaMA Guard 7B | Mean Perplexity (PPL)3.05 | 10 | |
| Jailbreaking | LLaMA Guard 2 8B | Bypass Rate100 | 1 | |
| Jailbreaking | LLaMA Guard 7B | Bypass Rate98.65 | 1 | |
| Latent Causal Mechanism Inference | Llama Guard LLM latent mechanism | Metric- | 0 |