INDEX
Explanations
mathematical properties
This neuron activates on technical terms naming convex‐geometry or functional‐analysis objects—words like “convex,” “normed,” “linear space,” “cone,” “closed,” “pointed,” “lattice,” and “polytope.”
New Auto-Interp
Negative Logits
Basket
-0.07
Listing
-0.07
Ngoài
-0.06
Ethiopian
-0.06
268
-0.06
ateş
-0.06
(Register
-0.06
πο
-0.06
reporter
-0.06
Aussie
-0.06
POSITIVE LOGITS
terra
0.07
abdom
0.07
quarterbacks
0.06
із
0.06
оди
0.06
ACCESS
0.06
hike
0.06
duğu
0.06
mach
0.06
!,
0.06
Activations Density 0.024%