INDEX

Explanations

keep your

The neuron identifies “keep”-style advice phrasing (instances of “keep” plus something to maintain or preserve).

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

𓋼

-1.34

 tornar

-1.31

ära

-1.30

 ocorrer

-1.28

 tocó

-1.27

 escolher

-1.24

 нема

-1.23

骓

-1.21

zegovina

-1.20

اوية

-1.20

POSITIVE LOGITS

and

1.62

或者

1.30

())),

1.27

해서

1.24

();

1.24

ㆆ

1.22

şi

1.20

 bestehen

1.17

ATION

1.16

 abreast

1.16

Activations Density 0.025%