INDEX

Explanations

numbers and identifiers

The neuron is highly sensitive to numeric tokens (multi‐digit numbers, years, IDs, etc.).

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

 skim

-1.50

 this

-1.49

 jelasnya

-1.48

ATIVO

-1.45

 olmak

-1.45

地道

-1.44

 этом

-1.41

icona

-1.40

iconductors

-1.40

imetsu

-1.39

POSITIVE LOGITS

謦

1.95

已经

1.80

鉺

1.71

 („

1.63

饪

1.61

䜣

1.58

1.52

熜

1.51

ﮦ

1.51

会

1.50

Activations Density 0.056%