list labels with colons

The neuron fires on items in structured “option” or “attribute‐value” lists—e.g. field names and their values (like “Colors:”, “White”, “Silver”, “Pink”, “Category:”, “Type:”, numeric codes, sizes, etc.).

New Auto-Interp

Configuration

Prompts (Dashboard)

392,802 prompts, 256 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Not in Any Lists

Negative Logits

industrie

0.56

atedral

0.56

agricultural

0.55

chandelier

0.54

aviation

0.54

établ

0.52

 अल्को

0.52

 किल्ल्या

0.51

Encycl

0.51

 الإنسان

0.51

POSITIVE LOGITS

 also

0.56

0.54

0.53

/*

0.53

0.52

 juga

0.52

Activations Density 0.303%