INDEX

Explanations

colors and items

The neuron fires primarily on tokens that appear in section or article headings (e.g. title‐case words, CSS class names used as labels, dates in headers).

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

-1.72

 también

-1.52

 such

-1.50

so

-1.44

 these

-1.41

it

-1.38

 augusti

-1.30

 своими

-1.30

umburg

-1.30

tal

-1.27

POSITIVE LOGITS

 мою

1.42

 colorear

1.38

 ordentlich

1.37

accessoires

1.31

klima

1.26

Colabor

1.25

 appea

1.22

ثم

1.22

 этого

1.19

emporio

1.19

Activations Density 0.117%