INDEX

Explanations

everything works

The neuron detects broad universal‐quantifier words such as “all,” “everything,” or their equivalents (e.g. Russian “все/всё”).

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

グルト

-1.55

 iſt

-1.53

ementara

-1.47

ากาศ

-1.47

纸

-1.42

 daca

-1.41

 Arrondissement

-1.41

 zijne

-1.40

 TOKYO

-1.40

if

-1.37

POSITIVE LOGITS

1.62

 será

1.52

You

1.40

Why

1.35

の人が

1.32

 incluindo

1.28

ра

1.26

甚至是

1.24

 universo

1.23

Activations Density 0.013%