INDEX
Explanations
quantitative measurements and specifications related to materials and their properties
New Auto-Interp
Neuron Alignment
Index
Value
% of L₁
326
+0.12
0.7%
382
+0.11
0.6%
83
+0.11
0.6%
Correlated Neurons
Index
P. Corr.
Cos Sim.
83
+0.12
0.04
342
+0.11
0.04
164
+0.11
0.05
Negative Logits
lando
-1.74
bugs
-1.59
secret
-1.55
admin
-1.52
ethics
-1.50
Anonymous
-1.49
Blog
-1.43
XML
-1.43
Microsoft
-1.42
Yahoo
-1.41
POSITIVE LOGITS
intervals
2.10
increments
2.01
+.
1.91
elapsed
1.90
+,
1.84
compared
1.84
depending
1.80
bps
1.78
averaged
1.78
epochs
1.76
Activations Density 0.335%