INDEX

Explanations

talking about lists of items

This neuron detects sentence-initial discourse markers or adverbial “filler” words (e.g. “Also,” “So,” “Actually,” “Of course,”) immediately followed by a comma.

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

aked

-1.44

 السياس

-1.44

 यह

-1.38

 betegs

-1.35

 rigue

-1.34

 fordi

-1.28

 mód

-1.27

vitational

-1.23

mbgg

-1.22

ѣмъ

-1.21

POSITIVE LOGITS

2.17

 people

1.94

 what

1.92

if

1.74

 think

1.68

how

1.62

to

1.59

 that

1.51

you

1.43

don

1.40

Activations Density 0.053%