INDEX
Explanations
People's names
The neuron fires on the components of personal names (especially surname fragments), i.e. it detects tokens that form people’s names.
New Auto-Interp
Negative Logits
Below
-0.06
cích
-0.06
Compare
-0.06
Directed
-0.06
borough
-0.06
Conditional
-0.06
secluded
-0.06
humour
-0.06
Attack
-0.06
converted
-0.06
POSITIVE LOGITS
改革
0.08
_videos
0.07
Env
0.07
.SubItems
0.07
resent
0.07
eighteen
0.07
ambient
0.06
现代
0.06
ображ
0.06
ियल
0.06
Activations Density 0.036%