INDEX
Explanations
mentions people and magazines
The neuron activates on proper names and titles—i.e. named entities like people’s names, organizations, and institutions.
New Auto-Interp
Negative Logits
morph
-0.07
ourd
-0.07
μετα
-0.07
تحلیل
-0.06
_in
-0.06
ستی
-0.06
buscar
-0.06
dir
-0.06
oidal
-0.06
============================================================================↵
-0.06
POSITIVE LOGITS
JSGlobal
0.07
หญ
0.07
_BROWSER
0.07
خرد
0.07
closet
0.06
mattresses
0.06
BaseEntity
0.06
CHKERRQ
0.06
-gray
0.06
Detective
0.06
Activations Density 0.057%