Stereotype Detection on SmallHeap (n=134)
79.9AccuracyOur detection & assessment step
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| Our detection & assessment stepEvaluation Protocol=Stereotype Detection & Assessment, t=0.692025.12 | 79.9 | 72 | 72.8 | 71.3 | |
| Our detection stepEvaluation Protocol=Stereotype Detection Only2025.12 | 68.7 | 65.6 | 75.1 | 67.5 | |
| Llama-Guard-3-8BEvaluation Protocol=Stereotype Detection Only2025.12 | 68.7 | 53.8 | 53.7 | 53.9 | |
| PerspectiveAPI IdentityAttackEvaluation Protocol=Stereotype Detection & Assessment, t=0.32025.12 | 67.2 | 63.6 | 71.7 | 65.2 | |
| ALBERT-V2Evaluation Protocol=Stereotype Detection Only2025.12 | 58.2 | 55.7 | 64.8 | 60.3 |