INDEX
Explanations
This neuron primarily activates on mentions of the “terahertz” (THz) frequency band or closely related subword tokens.
New Auto-Interp
Negative Logits
consequat
-0.07
>::
-0.07
nunca
-0.06
Anti
-0.06
Caption
-0.06
"]); ↵
-0.06
вокруг
-0.06
outbound
-0.06
Credit
-0.06
ิท
-0.06
POSITIVE LOGITS
discour
0.07
منزل
0.07
.hasMore
0.07
appeal
0.06
chantment
0.06
-conscious
0.06
GHz
0.06
'ils
0.06
UART
0.06
reminis
0.06
Activations Density 0.003%