INDEX

Explanations

and, than, except

The neuron strongly activates on base‐form action words (e.g. “verify,” “control,” “combat,” “treat,” “adjust,” “implement”).

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

to

-2.13

より

-1.58

 verschiedenen

-1.55

 двумя

-1.51

 bestimmten

-1.47

 возмо

-1.45

 ähnlichen

-1.45

of

-1.45

可能な

-1.43

なるべく

-1.43

POSITIVE LOGITS

and

2.03

 Péter

1.50

켭

1.45

ться

1.40

vnode

1.33

億円

1.33

 maît

1.28

.*/

1.27

 tedes

1.27

 remainder

1.25

Activations Density 0.025%