INDEX
Explanations
group, vulnerable, update, soul
the neuron activates on short technical identifiers/labels — e.g., code/package/entity names, acronyms, and UI/control identifiers.
New Auto-Interp
Negative Logits
ed
0.68
ä
0.61
വധി
0.55
譙
0.55
náv
0.55
QueryResult
0.54
াল
0.54
PlayerData
0.54
communicate
0.54
dataType
0.52
POSITIVE LOGITS
थे
0.53
ক
0.52
минима
0.51
interstices
0.51
शी
0.50
президент
0.50
authentication
0.49
이
0.49
Aspects
0.49
はじ
0.49
Activations Density 0.000%