INDEX

Explanations

numbers or punctuation followed by text

The neuron is most strongly triggered by structured metadata tokens—especially abbreviations with periods (like “U.S.”) and standalone numerals or date‐related digits.

New Auto-Interp

Configuration

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Not in Any Lists

Negative Logits

our

-1.59

 this

-1.52

私たち

-1.41

 that

-1.39

Sebagai

-1.32

 humeur

-1.31

𝕋

-1.27

рации

-1.26

 poule

-1.26

 Übersetzungen

-1.26

POSITIVE LOGITS

1.49

or

1.49

有何

1.45

 tomando

1.43

='';

1.40

 todav

1.34

stg

1.32

1.30

摒

1.29

 lacivert

1.27

Activations Density 1.181%