INDEX
Explanations
containers like buckets and baskets
New Auto-Interp
Neuron Alignment
Index
Value
% of L₁
1926
+0.13
0.4%
703
+0.12
0.4%
1013
+0.12
0.4%
Correlated Neurons
Index
P. Corr.
Cos Sim.
703
+0.13
0.04
1926
+0.12
0.04
289
+0.12
0.02
Negative Logits
Öster
-0.85
Præ
-0.84
stoff
-0.83
alkoh
-0.79
logis
-0.78
Spani
-0.77
Nö
-0.77
Rektor
-0.75
Okt
-0.75
hek
-0.75
POSITIVE LOGITS
containing
0.77
filled
0.66
container
0.62
containing
0.59
bucket
0.59
container
0.59
contains
0.58
bucket
0.57
feldspar
0.57
containers
0.56
Activations Density 0.226%