INDEX

Explanations

each and every one

The neuron specializes in spotting numeric values in structured metadata fields (e.g. form entries like phone numbers, ZIP codes, company IDs).

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

 citada

-1.45

 acoged

-1.43

 utilizada

-1.42

 reconocida

-1.41

 maravillosa

-1.38

 atractiva

-1.36

-1.32

warnai

-1.29

 enchufe

-1.27

女孩子

-1.24

POSITIVE LOGITS

每一

1.55

Any

1.54

This

1.41

 categorie

1.39

每一个

1.38

 każde

1.37

if

1.35

1.34

Each

1.34

每个

1.30

Activations Density 0.028%