INDEX
Explanations
This neuron detects the technical term “osteochondral” (and its subword components) in medical texts.
New Auto-Interp
Negative Logits
_fake
-0.08
-fed
-0.06
steal
-0.06
Mag
-0.06
"My
-0.06
optimizing
-0.06
Unauthorized
-0.06
〃
-0.06
.My
-0.06
drunk
-0.06
POSITIVE LOGITS
oth
0.07
ห
0.06
-cn
0.06
مول
0.06
ш
0.06
serge
0.06
대학
0.06
_OBJ
0.06
_OPTION
0.06
تن
0.06
Activations Density 0.002%