Gender Neuron Masking on InclusiveGender Masculine
1.39Avg. Gendered Terms per Response (M)LAPE
Evaluation Results
| Method | Links | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| LAPEBase Model=Llama 3.12026.05 | 1.39 | 0.18 | 0.25 | 11.57 | 1.48 | 2.1 | -1.65 | 0.59 | 0.42 | |
| sNeuron-TSTBase Model=Llama 3.12026.05 | 1.56 | 0.4 | 0.25 | 10.85 | 2.76 | 1.76 | -2.37 | 1.87 | 0.08 | |
| BaselineBase Model=Llama 3.12026.05 | 1.58 | 0.11 | 0.2 | 13.22 | 0.89 | 1.68 | — | — | — | |
| OursBase Model=Llama 3.12026.05 | 1.6 | 0.19 | 0.17 | 13.55 | 1.54 | 1.58 | 0.33 | -0.65 | -0.1 |