INDEX

Explanations

language suffixes

The neuron strongly activates on words and tokens that appear in section headings or list‐item titles (e.g. “točka,” “Svet,” numbered or titled points), so it’s detecting document structure markers—section or item 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

to

-1.35

Bety

-1.35

after

-1.34

Tils

-1.33

每個

-1.31

⦑

-1.30

無事に

-1.28

כד

-1.28

んび

-1.27

後で

-1.25

POSITIVE LOGITS

iramente

1.59

ophiles

1.50

ScopeManager

1.50

romeda

1.48

ceuticals

1.45

LAH

1.44

 '\\;'

1.43

:@"

1.42

ORUS

1.40

ğu

1.39

Activations Density 0.010%