INDEX
Explanations
Expressing thoughts or opinions
The neuron fires on introspective question phrases—especially the word “mind” in contexts like “on your mind”—i.e. when the assistant asks about the user’s thoughts or feelings.
New Auto-Interp
Negative Logits
(*)
-0.07
_allocation
-0.06
Ngh
-0.06
586
-0.06
Categoria
-0.06
immortal
-0.06
resurrect
-0.06
proper
-0.06
своїм
-0.06
265
-0.06
POSITIVE LOGITS
писание
0.06
γυνα
0.06
subscriptions
0.06
gon
0.06
打
0.06
Fed
0.06
olución
0.06
menin
0.06
وليو
0.06
طه
0.06
Activations Density 0.005%