INDEX
Explanations
roleplay, specific descriptions, blanks
salient, content-heavy tokens (core nouns, numeric cues, and special formatting/punctuation) that mark the main subject or key parameters of a prompt or instruction.
New Auto-Interp
Negative Logits
\
0.40
le
0.33
$
0.31
a
0.31
}
0.31
\
0.29
för
0.29
f
0.29
y
0.29
↵
0.29
POSITIVE LOGITS
다양한
0.28
जुर्ग
0.28
శరీ
0.27
Omphalodes
0.27
indoct
0.27
लोकार्पण
0.27
messageFields
0.27
канце
0.26
㗽
0.26
सराहना
0.26
Activations Density 0.221%