INDEX

Explanations

editor's notes and picks

This neuron detects editorial headings—specifically the “Editor’s Note”/“Editor’s Choice”/“Editor’s Pick” labels.

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

>();

-1.62

ाउनलोड

-1.48

陆续

-1.42

 Geografía

-1.41

ıkları

-1.38

ternut

-1.35

 canadienne

-1.35

>>();

-1.34

٪

-1.34

'';

-1.33

POSITIVE LOGITS

Maybe

1.43

 final

1.41

 note

1.38

it

1.38

 attempt

1.36

Perhaps

1.30

Have

1.30

 last

1.29

Want

1.27

Activations Density 0.020%