Gender Neuron Masking on GCGender Neutral
0Term Count (M) per ResponseBaseline
Evaluation Results
| Method | Links | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| BaselineBase Model=Llama 3.12026.05 | 0 | 0.01 | 1.69 | 0.01 | 0.06 | 10.12 | — | — | — | |
| OursBase Model=Llama 3.12026.05 | 0 | 0.01 | 1.68 | 0.01 | 0.06 | 10.13 | 0 | 0 | 0 | |
| LAPEBase Model=Llama 3.12026.05 | 0.01 | 0 | 1.84 | 0.04 | 0.03 | 9.66 | 0.02 | -0.03 | -0.46 | |
| sNeuron-TSTBase Model=Llama 3.12026.05 | 0.03 | 0.01 | 4.09 | 0.06 | 0.02 | 8.97 | 0.04 | -0.04 | -1.16 |