INDEX

Explanations

code and special characters

The neuron fires on code‐style identifiers used as keys or parameter names in assignments (e.g. “kernel”, “help”, “writer”, “ms”, “cat”).

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

and

-1.95

to

-1.25

can

-1.18

 found

-1.16

 hanya

-1.16

 coronav

-1.12

 tylko

-1.12

it

-1.12

 could

-1.11

 without

-1.11

POSITIVE LOGITS

一种

1.27

危机

1.26

dienne

1.25

still

1.23

if

1.23

 decorativos

1.22

分为

1.21

the

1.18

 úteis

1.16

 impresionantes

1.16

Activations Density 0.208%