INDEX

Explanations

country codes

This neuron is picking out lowercase, dot‐separated identifiers that look like Internet domain or package names (e.g. “ethz”, “edu”, “visualisation”).

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

Kepala

-1.46

 almohada

-1.45

有些不

-1.39

馓

-1.28

 inscrição

-1.27

astră

-1.27

 distinguer

-1.27

 münchen

-1.27

ݯ

-1.26

einf

-1.25

POSITIVE LOGITS

ariana

1.29

🚱

1.25

 dermed

1.19

enggarakan

1.18

 will

1.18

евна

1.18

 wraz

1.17

 bâtiment

1.17

 науки

1.15

 zejména

1.13

Activations Density 0.027%