INDEX
Explanations
Web addresses
The neuron strongly activates on marketing‐style or promotional text (e.g. “Looking to find…visit www…” calls to action, author‐byline blurbs, and similar ad-like language).
New Auto-Interp
Negative Logits
toned
-0.07
cz
-0.06
deletion
-0.06
opport
-0.06
.cp
-0.06
diagn
-0.06
inoc
-0.06
谓
-0.06
opol
-0.06
sinus
-0.06
POSITIVE LOGITS
IsActive
0.07
.Short
0.06
회사
0.06
takdir
0.06
UIAlert
0.06
_candidates
0.06
güçlü
0.06
CHECK
0.06
.period
0.06
Spectrum
0.06
Activations Density 0.021%