INDEX

Explanations

advice or prompts for action

This neuron detects call-to-action or directive phrases—imperative language urging the reader to do something (e.g. “don’t miss your chance,” “be sure to check out,” “listen carefully,” “keep watching,” etc.).

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

 however

-1.48

 However

-1.23

Huhu

-1.22

 همچنین

-1.20

 όμως

-1.10

 Thus

-1.09

 therefore

-1.05

いため

-1.05

However

-1.05

?><

-1.04

POSITIVE LOGITS

 your

1.54

 that

1.22

 đừng

1.21

 those

1.21

 there

1.20

it

1.19

!”

1.07

礫

1.05

 here

1.05

etc

1.04

Activations Density 0.029%