INDEX
Explanations
This neuron fires on verbs and phrases that offer or prompt ways to look up or find information—especially suggestions to “search,” “find,” “look up,” “try,” or go “online.”
New Auto-Interp
Negative Logits
コ
-0.07
ypad
-0.06
Symbols
-0.06
sorun
-0.06
ند
-0.06
تف
-0.06
(dl
-0.06
showError
-0.06
doğru
-0.06
contradiction
-0.06
POSITIVE LOGITS
GitHub
0.07
yní
0.07
*size
0.06
-len
0.06
-expression
0.06
]]
0.06
GE
0.06
Česk
0.06
sleeps
0.06
��
0.06
Activations Density 0.024%