INDEX
Explanations
The neuron signals strong matches for salient content-bearing nouns and named entities (important topic words) in the text.
New Auto-Interp
Negative Logits
(?)
0.47
などで
0.45
등으로
0.44
(?)
0.42
ではありません
0.42
𝖑
0.42
иногда
0.41
tetapi
0.41
등을
0.41
sauf
0.41
POSITIVE LOGITS
ซึ่ง
1.59
which
1.52
which
1.49
ซึ่ง
1.16
Which
1.14
WHICH
1.09
whiche
1.05
lequel
1.03
Which
1.02
który
1.02
Activations Density 0.248%