INDEX

Explanations

code comments and boilerplate

The neuron selectively activates on numeric tokens (i.e., number literals) in the code or comments.

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

んだよね

-1.49

 التاريخ

-1.39

 δεν

-1.33

もうすぐ

-1.33

 الخارجية

-1.31

 écoulé

-1.28

 Entdecken

-1.28

 baterie

-1.27

rógeno

-1.27

؟؟

-1.27

POSITIVE LOGITS

–

1.65

in

1.58

for

1.50

などは

1.36

to

1.31

なども

1.29

ce

1.21

がありますが

1.20

だが

1.18

by

1.17

Activations Density 0.001%