INDEX

Explanations

conventionally starts with a capital letter

The neuron detects numeric metadata (counts, dates, or metrics) in question‐and‐answer forum headers—e.g. numbers following “Answered,” “Replied,” “Posted,” 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

 загру

0.61

Batis

0.56

Much

0.53

Sur

0.52

"-------

0.52

ST

0.52

Sync

0.51

Jamaica

0.51

庸

0.50

 stö

0.50

POSITIVE LOGITS

boats

0.61

boat

0.57

 krav

0.55

 ultimo

0.54

 inzwischen

0.54

bottlecap

0.54

 вопро

0.53

 boats

0.52

ющи

0.51

टेन

0.51

Activations Density 0.001%