INDEX

Explanations

numbers in URLs

The neuron fires on numeric tokens (especially multi‐digit numbers) such as those in URLs or article/ID references.

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

 what

-1.46

 pflegen

-1.43

-1.38

may

-1.38

of

-1.34

 could

-1.29

-1.26

—

-1.26

 verbringen

-1.25

-1.22

POSITIVE LOGITS

了出去

1.45



1.41

шивания

1.39



1.38

)、

1.31

 personali

1.28

 vengono

1.28

 cowok

1.27

cally

1.25

好不好

1.23

Activations Density 0.177%