INDEX

Explanations

This invention

The neuron activates on numeric literals (digits and numbers) in code, e.g. the numeric values in the CSS.

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

it

-1.73

 will

-1.69

are

-1.59

-1.55

chsel

-1.53

);

-1.43

 炒め

-1.41

 čia

-1.38

tahui

-1.34

 beeind

-1.31

POSITIVE LOGITS

 बहुत

1.49

ícias

1.35

asje

1.27

 gangen

1.27

grained

1.27

owymi

1.27

atific

1.24

าน

1.24

drage

1.23

owym

1.21

Activations Density 0.003%