INDEX
Explanations
Asking questions
This neuron detects question‐related words and phrases—tokens that signal someone asking for information (e.g. ask, about, what, how, wonder).
New Auto-Interp
Negative Logits
King
-0.07
efficient
-0.07
bais
-0.06
stake
-0.06
renown
-0.06
_histogram
-0.06
Sound
-0.06
delegates
-0.06
Laud
-0.06
creepy
-0.06
POSITIVE LOGITS
Adopt
0.06
浙江
0.06
Росії
0.06
chtě
0.06
●●
0.06
...",↵
0.06
(<?
0.06
بدون
0.06
veled
0.05
_rl
0.05
Activations Density 0.037%