INDEX

Explanations

hypothetical vs real-world scenarios

New Auto-Interp

Configuration

Prompts (Dashboard)

392,802 prompts, 256 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 compressor

1.25

 или

1.23

or

1.16

然后在

1.16

 Владимир

1.14

akam

1.13

或者

1.13

FormComponent

1.10

₁,

1.09

but

1.08

POSITIVE LOGITS

ting

1.42

 crucially

1.39

 offenbar

1.37

 stubbornly

1.35

weist

1.34

 nonetheless

1.32

 మాత్రం

1.30

 persists

1.29

 shattered

1.27

 shrewd

1.27

Activations Density 2.187%