INDEX

Explanations

wish to

The neuron detects polite advisory or permission phrases (modals and soft suggestions like “may,” “wish,” “choose,” “desire,” followed by “to”).

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

are

-1.79

The

-1.64

-1.58

-1.56

-1.50

 essentially

-1.48

都能

-1.45

 four

-1.40

//

-1.39

He

-1.34

POSITIVE LOGITS

ومن

1.51

 princípios

1.34

پول

1.34

 conoscere

1.32

 alberi

1.32

ᐯ

1.28

渽

1.26

 pią

1.25

 Transformer

1.25

 preços

1.25

Activations Density 0.018%