INDEX
Explanations
lists of letters and subsequent word parts
the neuron detects isolated single-character tokens (single letters or initials) across scripts.
New Auto-Interp
Negative Logits
ција
0.17
اداس
0.17
णिमा
0.17
轳
0.17
সম্পাদকীয়
0.16
መሳ
0.16
ഹം
0.16
mauvais
0.16
াদেশিক
0.16
distanceArray
0.16
POSITIVE LOGITS
a
0.26
A
0.24
I
0.24
E
0.21
n
0.21
t
0.21
O
0.19
l
0.19
X
0.18
Q
0.18
Activations Density 0.053%