INDEX

Explanations

phrases starting with certain letter combinations

The neuron activates strongly on short character fragments that are parts of words broken by hyphens or line‐break hyphenation, i.e. subword pieces resulting from split words.

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

 only

-1.79

-1.55

 They

-1.51

 these

-1.48

 there

-1.40

↵↵

-1.36

 also

-1.35

 always

-1.32

やはり

-1.29

他にも

-1.27

POSITIVE LOGITS

 dátum

1.53

 ucran

1.52

coden

1.50

 sinds

1.45

ܖ

1.45

 nór

1.43

 kateg

1.42

 fidé

1.41

ndash

1.41

 rév

1.38

Activations Density 0.164%