INDEX

Explanations

image, text, object, series, credit card

This neuron detects Japanese katakana sequences (i.e. loanwords and technical terms written in katakana).

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

for

-1.63

may

-1.52

or

-1.36

\\

-1.34

on

-1.34

 either

-1.34

-1.32

-1.30

it

-1.29

in

-1.27

POSITIVE LOGITS

 durer

1.63

 célè

1.55

Viitteet

1.48

 schermo

1.46

cellona

1.44

 pozi

1.43

 ранее

1.39

новения

1.39

ження

1.38

摞

1.38

Activations Density 0.004%