INDEX
Negative Logits
and
1.04
or
1.03
on
0.97
is
0.91
but
0.91
has
0.88
by
0.79
r
0.78
et
0.77
他
0.77
POSITIVE LOGITS
u
1.16
ו
1.16
erode
1.05
is
1.02
ку
0.97
ра
0.95
are
0.95
the
0.94
p
0.94
t
0.93
Activations Density 0.001%
and
or
on
is
but
has
by
r
et
他
u
ו
erode
is
ку
ра
are
the
p
t