INDEX

Explanations

go and do things

This neuron strongly activates on imperative (“command”) action verbs (e.g. “get,” “find,” “go,” “talk,” “experience,” “party,” “beat,” “blog”).

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

or

-1.45

带领

-1.39

 conectados

-1.37

 bekende

-1.36

And

-1.30

 MUCH

-1.28

 genaue

-1.26

 lengte

-1.26

 wichtigste

-1.26

 each

-1.25

POSITIVE LOGITS

 サービス

1.51

roffene

1.50

 fluff

1.44



1.44

海の

1.43

zości

1.41

Partager

1.41

וּ

1.41

ENDIF

1.39

 palsu

1.38

Activations Density 0.007%