INDEX
Explanations
news articles
The neuron fires strongly on named entities and their immediately following keywords (e.g. proper nouns like product or place names and their action verbs) as they appear in lead‐off positions of news‐style sentences.
New Auto-Interp
Negative Logits
tatto
-0.07
�
-0.06
ALS
-0.06
_dc
-0.06
numpy
-0.06
kü
-0.06
kla
-0.06
Mapper
-0.06
Dover
-0.06
plagued
-0.06
POSITIVE LOGITS
絶
0.08
seaside
0.07
ají
0.07
شناسی
0.06
배열
0.06
prop
0.06
.margin
0.06
_ERROR
0.06
Hành
0.06
weblog
0.06
Activations Density 0.186%