INDEX
Explanations
descriptions related to a cookbook or recipes
the neuron detects named entities — especially organization, brand, company, and country names (proper nouns).
New Auto-Interp
Neuron Alignment
Index
Value
% of L₁
1343
+0.14
0.4%
856
+0.13
0.4%
50
+0.13
0.4%
Correlated Neurons
Index
P. Corr.
Cos Sim.
856
+0.14
0.05
1806
+0.13
0.05
16
+0.13
0.05
Negative Logits
ностран
-0.62
<bos>
-0.60
jws
-0.54
checs
-0.54
спользова
-0.53
споль
-0.53
://"
-0.50
دیکھیے
-0.50
dasselbe
-0.50
synes
-0.48
POSITIVE LOGITS
stockholm
1.06
accla
1.06
franz
1.00
Departement
0.97
inev
0.97
casio
0.96
daf
0.95
intermitt
0.95
Cfr
0.95
ivi
0.94
Activations Density 0.238%