INDEX
Explanations
YouTube videos
The neuron is especially sensitive to tokens used in enumerated response lists or tutorial‐style step headings (e.g. numbers, “Getting,” “Started,” “with,” etc.), effectively flagging list‐item and instructional‐heading language.
New Auto-Interp
Negative Logits
_variant
-0.07
avi
-0.07
educators
-0.07
Recording
-0.06
Translator
-0.06
(video
-0.06
문
-0.06
悟
-0.06
/****************************************************************
-0.06
irect
-0.06
POSITIVE LOGITS
alers
0.07
syrup
0.06
peng
0.06
prim
0.06
PIT
0.06
hashes
0.06
trailing
0.06
schnell
0.06
هوش
0.06
provoz
0.06
Activations Density 0.036%