INDEX

Explanations

code structure imports

The neuron is specialized to detect numeric literals and digit-containing tokens (i.e. numbers or identifiers with numeric parts).

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

どうやら

-1.54

 একটি

-1.47

他說

-1.41

我が家の

-1.41

 используется

-1.39

黉

-1.38

檗

-1.34

↓↓

-1.33

izador

-1.30

なぜか

-1.30

POSITIVE LOGITS

 with

1.63

 their

1.62

all

1.55

any

1.52

 would

1.52

 could

1.51

 success

1.45

 from

1.44

leiding

1.37

out

1.34

Activations Density 0.029%