INDEX

Explanations

number followed by one or nine

The neuron fires on explicit numeric tokens (e.g. digits, list‐position numbers, “number one” counts).

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

for

-2.08

of

-1.68

鉛筆

-1.63

-1.59

 ziemlich

-1.51

by

-1.50

 passieren

-1.48

 following

-1.47

独特

-1.46

or

-1.45

POSITIVE LOGITS

 akces

1.70

٣

1.63

hrten

1.55

RÉ

1.54

翆

1.52

 wortel

1.48

spodnie

1.48

 reager

1.47

ieteur

1.46

埽

1.45

Activations Density 0.024%