INDEX

Explanations

return, throw, if

The neuron activates on runs of leading whitespace (indentation) in code blocks.

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

'./

-1.56

 هذا

-1.52

 SALUD

-1.41

絝

-1.41

 pourront

-1.38

 regalías

-1.38

緲

-1.38

 this

-1.37

 MINISTRY

-1.32

 FRENCH

-1.31

POSITIVE LOGITS

 Before

1.43

is

1.37

end

1.33

1.32

id

1.32

 when

1.31

 cumbersome

1.25

1.24

Activations Density 0.006%