INDEX
Explanations
technology
The neuron activates on descriptive adjectives (especially those highlighting features or qualities, like “organic,” “curved,” “3-D,” “Islamic,” etc.).
New Auto-Interp
Negative Logits
スカ
-0.08
Mild
-0.07
wiped
-0.07
Attention
-0.07
reiterated
-0.06
次
-0.06
roaming
-0.06
JJ
-0.06
.design
-0.06
iples
-0.06
POSITIVE LOGITS
است
0.07
_ctr
0.07
ین
0.06
posicion
0.06
Unicode
0.06
hindsight
0.06
$MESS
0.06
spi
0.06
τικών
0.06
"';
0.06
Activations Density 0.163%