INDEX

Explanations

might followed by a verb

New Auto-Interp

Configuration

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

of

-2.06

鯵

-1.38

 ‘’

-1.27

として

-1.24

 Ganzen

-1.23

 avancer

-1.21

рони

-1.20

 δὲ

-1.17

dym

-1.17

 exprimer

-1.14

POSITIVE LOGITS

 there

1.75

 while

1.66

because

1.40

two

1.38

حات

1.28

if

1.24

 most

1.23

 less

1.23

もしれない

1.23

it

1.22

Activations Density 0.070%