INDEX

Explanations

The neuron is picking out capitalized code identifiers (e.g. class/type names) in programming text.

shebang lines and code constructs

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

我一

-1.29

-1.25

 wiederhol

-1.24

少なくとも

-1.17

其它

-1.17

面白かった

-1.16

 erfolgreiche

-1.15

Other

-1.09

 untuk

-1.09

 most

-1.07

POSITIVE LOGITS

1.47

 theſe

1.20

那个

1.16

却

1.14

 SUCH

1.13

 queste

1.13

 νέ

1.13

<bos>

1.13

 ベー

1.12

 vrouwen

1.11

Activations Density 0.000%