INDEX
Explanations
existing methods
This neuron detects words signaling existing or prior‐art technology (e.g. “existing,” “current,” “typical,” “conventional”).
New Auto-Interp
Negative Logits
col
-0.07
دشمن
-0.07
Ferr
-0.06
刘
-0.06
ocard
-0.06
lent
-0.06
Lou
-0.06
manganese
-0.06
Curt
-0.06
Had
-0.06
POSITIVE LOGITS
glfw
0.07
('''0.07
Maritime
0.06
//}}
0.06
cambio
0.06
Serialization
0.06
onda
0.06
президент
0.06
Console
0.06
XM
0.06
Activations Density 0.023%