INDEX
Explanations
animal traits: intelligent, loyal, trainable
This neuron detects formatting and structural markup in the text (headings, emphasis/bold markers, section bullets and similar layout tokens).
New Auto-Interp
Negative Logits
русский
0.44
обы
0.41
گستر
0.39
питание
0.39
ствует
0.38
действует
0.38
unidentified
0.38
IONES
0.38
vét
0.38
formulas
0.37
POSITIVE LOGITS
affectionate
0.74
loyal
0.72
trainable
0.70
intelligent
0.66
companionship
0.63
Intelligent
0.62
loyalty
0.62
docile
0.61
Loyal
0.59
inteligentes
0.58
Activations Density 0.035%