INDEX

Explanations

numbers and years

The neuron fires on tokens that look like fragments of long alphanumeric codes (e.g. Base64‐style or ID strings).

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

𝗜

-1.51

你有

-1.49

$}

-1.41

México

-1.41

jahre

-1.39

หาก

-1.34

ถ้า

-1.33

mıştır

-1.32

mmmm

-1.30

wanderung

-1.30

POSITIVE LOGITS

崭

1.70

蓠

1.63

borderSide

1.61

摞

1.55

 this

1.53

 jälkeen

1.43

>-

1.41

 idyllic

1.39

 most

1.38

rebbe

1.37

Activations Density 0.022%