INDEX

Explanations

email domain names

The neuron fires on top‐level domain tokens (like “com,” “us,” “edu,” “gov”) in email addresses or URLs.

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

 charming

-1.50

my

-1.46

 instala

-1.45

瞥

-1.41

 ajudam

-1.39

 batalha

-1.38

 língua

-1.38

 picturesque

-1.37

 peuvent

-1.37

اقتصاد

-1.36

POSITIVE LOGITS

 estekak

1.59

entown

1.38

 kvě

1.36

or

1.34

 ścian

1.33

开始

1.30

ってました

1.30

Vendredi

1.30

gown

1.29

We

1.26

Activations Density 0.020%