INDEX

Explanations

CR followedby Rail or item codes

The neuron fires on short alphanumeric labels and codes—uppercase acronyms or identifiers (often including digits and hyphens) such as “F25,” “CRPD-248,” “Spoiler,” or “Read More.”

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

création

-1.62

 techni

-1.57

跸

-1.51

ᑲ

-1.50

 dormitorios

-1.46

*:

-1.44

撄

-1.41

℠

-1.40

ﾠ

-1.38

‹

-1.37

POSITIVE LOGITS

all

1.55

 with

1.54

 including

1.39

BIÉN

1.37

 nieruch

1.35

gonic

1.35

 meses

1.34

 each

1.33

 start

1.33

 That

1.32

Activations Density 0.001%