INDEX

Explanations

convert data or text

The neuron is picking out domain-specific technical or content-bearing terms (e.g. “real,” “life,” “game,” “print,” “data,” “import,” “<,” “div,” “toxic,” etc.).

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

to

-1.65

 each

-1.46

 olmak

-1.45

 these

-1.41

 dziś

-1.36

 this

-1.34

it

-1.30

 persönlichen

-1.27

 Möglichkeit

-1.25

 from

-1.23

POSITIVE LOGITS

oció

1.55

躂

1.49

izacao

1.42

而不是

1.37

 instead

1.37

 вместо

1.34

 حتی

1.31

tyfik

1.29

ଃ

1.28

 sadece

1.27

Activations Density 0.147%