| Task Name | Dataset Name | SOTA Result | Trend | |
|---|---|---|---|---|
| Refusal | Refusal benchmark | Alignment Probability97.81 | 55 | |
| Safety Refusal | Refusal | ASR0 | 42 | |
| Safety Evaluation | Refusal | Refusal Rate86.6 | 36 | |
| Refusal Behavior | Refusal | Sorry Rate69.3 | 22 | |
| Refusal | Refusal (test) | BAS (Baseline)64.3 | 6 | |
| Concept Recovery | REFUSAL | Mean MCC0.5669 | 6 | |
| Concept Identifiability | Refusal | MCC0.9 | 6 |