INDEX
Explanations
towns and villages
This neuron detects the repeated article title line (a place name) at the very start of Wikipedia‐style settlement entries, just before the opening parenthesis.
New Auto-Interp
Negative Logits
HU
-0.07
builder
-0.06
.are
-0.06
bathtub
-0.06
emoji
-0.06
sad
-0.06
Brand
-0.06
afari
-0.06
worm
-0.06
(now
-0.06
POSITIVE LOGITS
idades
0.06
quo
0.06
家庭
0.06
располож
0.06
<Contact
0.06
//================================================================
0.06
бор
0.06
경우
0.06
پیشینه
0.06
_contact
0.06
Activations Density 0.004%