INDEX

Explanations

Halle Hallelujah

This neuron fires on standalone tokens that start lines or sections—e.g. the first word of a lyric line, a chord label in a song intro, or “Re” in legal-citation 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

but

-1.45

and

-1.44

 then

-1.36

などなど

-1.30

in

-1.30

It

-1.29

だけど

-1.26

-1.24

也会

-1.24

也是

-1.21

POSITIVE LOGITS

 mogli

1.79

Ukuran

1.63

Kelebihan

1.55

以來

1.53

糴

1.49

Karakteristik

1.45

 académico

1.45

our

1.45

苙

1.43

何とか

1.42

Activations Density 0.171%