INDEX

Explanations

describes properties or roles

The neuron fires on high‐frequency multiword chunks—common collocations such as “well known,” “were known as,” “ability to,” or “a lot of.”

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

 this

-1.39

it

-1.35

 before

-1.15

 there

-1.14

 warum

-1.13

).

-1.13

 which

-1.13

 them

-1.10

 they

-1.09

 Dicapai

-1.09

POSITIVE LOGITS

 also

1.66

 även

1.38

 også

1.38

 fré

1.36

 myös

1.34

 fenti

1.32

 inoltre

1.30

 också

1.27

 Père

1.23

Also

1.23

Activations Density 0.074%