INDEX
Explanations
multilingual suffixes and conjunctions
New Auto-Interp
Negative Logits
Software
0.64
totaling
0.62
software
0.59
ed
0.57
building
0.56
と呼ばれる
0.55
Building
0.55
which
0.54
amounting
0.53
based
0.53
POSITIVE LOGITS
גם
0.78
себе
0.77
też
0.76
также
0.70
aussi
0.69
się
0.68
પણ
0.68
শুধু
0.66
сначала
0.66
కూడా
0.66
Activations Density 0.141%