INDEX

Explanations

very + noun

This neuron detects emphatic “the very … of …” constructions, i.e. noun phrases framed by “the very [noun] of [noun].”

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

 seulement

-1.82

also

-1.66

叆

-1.65

at

-1.62

 également

-1.60

 tudi

-1.59

 אנשים

-1.52

 also

-1.52

and

-1.49

และ

-1.47

POSITIVE LOGITS

卨

1.66

都不会

1.63

 того

1.63

・・・・

1.58

當

1.57

aughty

1.52

ﾍ

1.52

.•

1.50

 küche

1.49

最後

1.47

Activations Density 0.006%