INDEX

Explanations

calculations and reasons

The neuron is strongly activated by two-digit numeric tokens (e.g. reference numbers, section IDs, version numbers).

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

ありますか

-1.41

是怎么回事

-1.34

 corredores

-1.30

 concursos

-1.29

not

-1.27

 with

-1.27

店で

-1.27

なった

-1.26

始め

-1.23

なくても

-1.22

POSITIVE LOGITS

 زیرا

1.41

[]

1.33

 meisten

1.32

 GERMANY

1.31

<",

1.28

 поскольку

1.27

 ITALY

1.24

()");

1.24

"',

1.23

 kerana

1.23

Activations Density 0.013%