INDEX

Explanations

adding or creating

The neuron fires on relatively uncommon, content-rich “keyword” tokens (i.e. the informative nouns and verbs that carry the main semantic load of a passage).

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

-1.96

 then

-1.64

 just

-1.43

 especially

-1.42

 more

-1.30

so

-1.30

especially

-1.27

 chiave

-1.23

 так

-1.23

 even

-1.21

POSITIVE LOGITS

 jednou

1.59

one

1.52

 некоторы

1.49

 that

1.40

 aquela

1.30

yaitu

1.27

 olhando

1.27

seperti

1.27

要么

1.26

мання

1.26

Activations Density 0.078%