INDEX

Explanations

designed to

The neuron strongly activates on prominent action verbs that describe usage, function, or benefits (e.g. “used,” “save,” “receive,” “drive,” “manage”).

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

in

-1.73

to

-1.63

 Also

-1.61

For

-1.55

 required

-1.51

 more

-1.47

or

-1.45

for

-1.45

 commonly

-1.43

on

-1.37

POSITIVE LOGITS

intended

1.41

certainly

1.40

 própri

1.34

recently

1.34

probably

1.34

 colonnes

1.33

뭄

1.33

 siè

1.33

quite

1.32

只有

1.31

Activations Density 0.147%