INDEX

Explanations

various languages and academic topics

The neuron is strongly activated by numeric tokens—especially page, section, or chapter numbers (including decimals) in things like tables of contents.

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

 when

-1.76

-1.59

 which

-1.57

coltà

-1.55

 kapsul

-1.54

homonymie

-1.53

criminate

-1.52

those

-1.51

anız

-1.48

 mene

-1.46

POSITIVE LOGITS

เพื่อน

1.28

 laterales

1.28

 більш

1.21

گوشی

1.21

ется

1.20

–

1.16

 взял

1.16

次は

1.15

 kaksi

1.15

“

1.13

Activations Density 0.002%