INDEX
Explanations
months and days
The neuron activates on month names in date phrases (e.g. “Last July,” “Last January,” etc.), effectively spotting temporal references.
New Auto-Interp
Negative Logits
�
-0.06
OSE
-0.06
Gry
-0.06
stream
-0.06
black
-0.06
Kral
-0.06
simulated
-0.06
postpone
-0.06
spouse
-0.06
ترکی
-0.06
POSITIVE LOGITS
защиты
0.07
еса
0.06
.listen
0.06
příslu
0.06
ือน
0.06
susceptible
0.06
زش
0.06
.ALIGN
0.06
základ
0.06
don
0.06
Activations Density 0.013%