INDEX

Explanations

technical and engineering documents

Looking at the activations, this neuron responds strongly to: - Whitespace/formatting characters (tabs, spaces, line breaks) - Punctuation marks (periods, commas) - Parts of email addresses and URLs (especially periods and @ symbols) - Numbers in

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

Neden

-1.49

ープン

-1.48

cribes

-1.46

Soorten

-1.45

источник

-1.41

Mvh

-1.39

Татьяна

-1.37

Ար

-1.36

Begriff

-1.34

usions

-1.31

POSITIVE LOGITS

 also

1.78

 which

1.66

of

1.59

and

1.33

you

1.31

/**

1.30

就要

1.30

燉

1.27

 equip

1.24

 ему

1.22

Activations Density 0.409%