INDEX

Explanations

protection from damage and attack

The neuron fires strongly on numeric expressions (especially multi‐digit and decimal numbers) indicating quantities or measurements.

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

in

-1.94

to

-1.84

 speziellen

-1.71

for

-1.47

 persönlichen

-1.46

 outras

-1.40

 zusätzlichen

-1.38

 verstär

-1.38

これからの

-1.36

すぎた

-1.36

POSITIVE LOGITS

۶

1.47

高

1.29

1.28

 both

1.27

any

1.26

۴

1.26

all

1.26

۵

1.25

ｋ

1.24

Activations Density 0.076%