INDEX

Explanations

specific named entities and terms

The neuron fires on section headings or metadata titles—usually uppercase or title‐case words marking headers (like “Vol,” “No,” “Urgent,” “Trigger,” “MAGENTO Catalog,” 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

of

-1.84

咘

-1.52

();

-1.39

bluza

-1.37

 KIDS

-1.33

dáms

-1.33

 andere

-1.33

opération

-1.32



-1.31

nomor

-1.30

POSITIVE LOGITS

 because

1.75

 that

1.61

 هایی

1.55

maline

1.53

 when

1.52

1.50

Two

1.48

That

1.47

ogados

1.42

ldots

1.41

Activations Density 0.279%