INDEX
Explanations
Conversational text
The neuron strongly activates on adverbs and small transitional or discourse‐marker words (e.g. “just,” “quickly,” “here,” “in the meantime”) that link or steer the flow of a text.
New Auto-Interp
Negative Logits
rit
-0.07
suç
-0.06
dripping
-0.06
svých
-0.06
293
-0.06
Seen
-0.06
_fixture
-0.05
[ix
-0.05
등학교
-0.05
頭
-0.05
POSITIVE LOGITS
independents
0.08
dül
0.08
зберіг
0.07
Throwable
0.07
.TEXT
0.07
biệt
0.07
BigInteger
0.07
-responsive
0.07
Propel
0.07
GLfloat
0.07
Activations Density 0.098%