Stereotype Detection on StereoSet (n=237)
93.4AccuracyOur detection step
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| Our detection stepEvaluation Protocol=Stereotype Detection Only2025.12 | 93.4 | 91.9 | 91.4 | 92.4 | |
| ALBERT-V2Evaluation Protocol=Stereotype Detection Only2025.12 | 93.4 | 91.4 | 88.9 | 95.2 | |
| Our detection & assessment stepEvaluation Protocol=Stereotype Detection & Assessment, t=02025.12 | 93.1 | 91.4 | 90.5 | 92.4 | |
| PerspectiveAPI IdentityAttackEvaluation Protocol=Stereotype Detection & Assessment, t=0.252025.12 | 83.7 | 77.3 | 74.6 | 84.1 | |
| Llama-Guard-3-8BEvaluation Protocol=Stereotype Detection Only2025.12 | 76.5 | 60.7 | 60.5 | 79.3 |