Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

Neuron Labeling on SAE Vanilla (evaluated neurons)

0.8AUC

CSP

0.73760.75380.770.7862Apr 24, 2026
Updated 1mo ago

Evaluation Results

MethodLinks
2026.04
0.8
2026.04
0.8
2026.04
0.78
2026.04
0.78
2026.04
0.78
2026.04
0.77
2026.04
0.77
2026.04
0.77
2026.04
0.76
2026.04
0.76
2026.04
0.76
2026.04
0.76
2026.04
0.75
2026.04
0.75
2026.04
0.74