INDEX
Explanations
high importance
This neuron responds to evaluative and intensifying words—adjectives and adverbs that mark emphasis or promotion (e.g. “most,” “promising,” “greatly,” “urgently”).
New Auto-Interp
Negative Logits
giờ
-0.08
wo
-0.07
-0.07
难
-0.07
-0.06
Bit
-0.06
dictates
-0.06
endregion
-0.06
안
-0.06
-0.06
POSITIVE LOGITS
갤로그
0.06
offsetX
0.06
nutné
0.06
''}↵
0.06
nova
0.06
))}↵
0.06
das
0.06
)>↵
0.06
이동
0.06
Nová
0.06
Activations Density 0.059%