INDEX

Explanations

technical file paths and errors

The neuron fires most strongly on named‐entity–style tokens (e.g. document headings, proper names, file‐ or function‐names) rather than ordinary running text.

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

 çünkü

-2.45

 такие

-1.54

 chré

-1.53

kimono

-1.52

contoh

-1.51

alami

-1.50

gamis

-1.48

 दुनिया

-1.47

our

-1.45

 нашей

-1.44

POSITIVE LOGITS

ַּ

2.06

ּוֹ

2.05

ֶּ

1.92

oted

1.90

vább

1.81

 تقری

1.81

are

1.76

 tersebut

1.68

ment

1.60

haviours

1.60

Activations Density 0.000%