INDEX

Explanations

call and challenge

The neuron detects phrases that signal a challenge or call to action (e.g. “laid down the gauntlet,” “call to action,” “challenge”).

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

şi

-1.51

 elástica

-1.32

 sintética

-1.31

 vacunas

-1.30

sprogram

-1.26

 frutto

-1.24

嚴重

-1.23

 farbe

-1.21

»)

-1.21

iramente

-1.20

POSITIVE LOGITS

</b>

1.42

berti

1.35

がいる

1.25

hynch

1.24

to

1.24

“

1.21

’

1.21

ご購入

1.16

こちらも

1.15

venger

1.15

Activations Density 0.047%