INDEX
Explanations
the sequence "<bos>" marking the beginning of a text
New Auto-Interp
Neuron Alignment
Index
Value
% of L₁
674
+0.44
7.1%
1741
+0.07
1.0%
1120
+0.04
0.7%
Correlated Neurons
Index
P. Corr.
Cos Sim.
674
+0.44
0.92
823
+0.07
0.93
654
+0.04
0.94
Negative Logits
ে
-1.13
भी
-1.12
не
-1.12
ко
-1.10
া
-1.10
про
-1.09
но
-1.09
所有
-1.09
ার
-1.09
الف
-1.08
POSITIVE LOGITS
<bos>
13.11
encomp
4.75
fuf
4.73
guarante
4.68
affor
4.67
effe
4.66
squa
4.65
fta
4.62
increa
4.61
secon
4.58
Activations Density 0.783%