INDEX
Explanations
Image/figure references in research papers
This neuron never activates on any token—it appears to be effectively “dead” and does not detect any pattern.
New Auto-Interp
Negative Logits
Position
-0.06
ues
-0.06
Arial
-0.06
Split
-0.06
ustain
-0.06
Distance
-0.06
########################################################
-0.05
>I
-0.05
shattered
-0.05
-0.05
POSITIVE LOGITS
specifier
0.08
للإ
0.07
شما
0.07
القي
0.06
NSCoder
0.06
över
0.06
connecting
0.06
cido
0.06
(eventName
0.06
cumbersome
0.06
Activations Density 0.001%