INDEX

Explanations

listing specific examples after "include:"

The neuron fires on list‐introduction phrases announcing “Notable people … include,” i.e. the word “include” in those headers.

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

of

-2.41

The

-1.79

 This

-1.69

-1.52

to

-1.46

-1.45

mi

-1.45

in

-1.43

You

-1.41

an

-1.40

POSITIVE LOGITS

 følgende

1.71

1.66

 kalori

1.63

 abiye

1.63

 maksi

1.58

 diki

1.56

Makanan

1.55

opheles

1.53

 kemer

1.53

 plafon

1.52

Activations Density 0.092%