INDEX

Explanations

wanting to

The neuron flags high–information content words (key nouns and verbs) rather than common function words.

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

猻

-1.31

௅

-1.30

 poète

-1.27

 společnost

-1.26

签订

-1.21

 huelga

-1.20

Tweets

-1.19

 douane

-1.18

 aplikace

-1.17

 circonst

-1.16

POSITIVE LOGITS

 then

1.55

 about

1.40

}_{

1.34

 taken

1.34

那么

1.30

",

1.27

\}\

1.22

1.18

 verknüp

1.17

↵↵↵

1.16

Activations Density 0.152%