INDEX

Explanations

plan, developer, fashion, commissioner, office

The neuron fires on multi‐word, capitalized key terms—like program or policy titles, brand names, organization or job titles, and person names—essentially flagging important named entities and headings.

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

 então

-1.51

 chiare

-1.45

思議

-1.43

 тогда

-1.40

 sprawie

-1.37

numericUpDown

-1.32

罈

-1.32

for

-1.32

in

-1.31

groupBox

-1.31

POSITIVE LOGITS

AfterViewInit

1.63

1.62

aneers

1.55

átku

1.52

and

1.51

reurs

1.48

Its

1.42

its

1.37

 JULY

1.34

шения

1.32

Activations Density 0.119%