INDEX

Explanations

. followed by new sentence start

The neuron strongly activates on tokens that mark the beginnings of lines or sentences—e.g. leading whitespace or capitalized, sentence‐initial words.

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

 például

-1.35

久し

-1.33

 embalaje

-1.32

 дори

-1.27

 célè

-1.27

 چهارم

-1.25

 stratég

-1.24

んだろう

-1.23

 increíbles

-1.23

 човек

-1.23

POSITIVE LOGITS

 only

1.57

 there

1.51

 doesn

1.38

 before

1.33

it

1.32

put

1.32

 when

1.30

go

1.27

 because

1.27

 without

1.24

Activations Density 0.027%