The TPR-GAP results for all versions (Balanced, Gentle, and Aggressive) for the three concepts where we have ground truth (Adjectives, Gender, and Race).
Experiment . | Adjectives . | Gender . | Race . |
---|---|---|---|
Balanced | 0.057 | 0.003 | 0.002 |
Gentle | 0.074 | 0.014 | 0.012 |
Aggressive | 0.012 | 0.003 | 0.049 |
Experiment . | Adjectives . | Gender . | Race . |
---|---|---|---|
Balanced | 0.057 | 0.003 | 0.002 |
Gentle | 0.074 | 0.014 | 0.012 |
Aggressive | 0.012 | 0.003 | 0.049 |