INDEX

Explanations

This neuron is tuned to recognize nouns ending in common derivational suffixes (e.g. –ers, –ion, –ment).

specific topics or subjects

New Auto-Interp

Configuration

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Negative Logits

and

-2.19

gnition

-1.49

alain

-1.40

chaise

-1.37

cardigan

-1.36

of

-1.33

It

-1.32

gamis

-1.29

bantu

-1.28

sopa

-1.28

POSITIVE LOGITS

Ī

1.22

–

1.16

izat

1.13

eping

1.13

 sige

1.13

oping

1.07

 ruine

1.07

豋

1.06

 beri

1.06

 keten

1.05

Activations Density 0.100%