INDEX

Explanations

domain names and email addresses

The neuron is picking up the little “at”/“dot” tokens that pepper e-mail addresses and URLs in text.

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

and

-1.45

 when

-1.43

as

-1.42

：</

-1.42

 using

-1.40

it

-1.28

 after

-1.27

 before

-1.26

 его

-1.25

 making

-1.24

POSITIVE LOGITS

mainly

1.58

either

1.52

besides

1.48

⁛

1.48

鼱

1.45

already

1.41

 permettent

1.39

appears

1.38

 consigue

1.38

anyway

1.34

Activations Density 0.007%