INDEX

Explanations

package names and identifiers

This neuron detects two-letter lowercase identifiers (like “cz,” “de,” “ch,” etc.) that appear as segments in package or import paths (i.e., country‐code TLD style package prefixes).

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

but

-1.57

 without

-1.41

 because

-1.40

on

-1.37

if

-1.34

out

-1.31

all

-1.28

not

-1.21

one

-1.20

by

-1.17

POSITIVE LOGITS

 $_{\

1.22

 decidedly

1.18

 $^{\

1.16

ngths

1.15

 delightfully

1.15

 majestic

1.14

$-$

1.14

 umgangssprach

1.13

 quintessential

1.13

 impromptu

1.11

Activations Density 0.008%