INDEX

Explanations

code functions and variables

The neuron primarily detects opening curly‐brace tokens (“{”) that mark the start of 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

 campesinos

-1.38

lump

-1.19

 vorsichtig

-1.16

 Proceed

-1.11

 textu

-1.10

 vaqueros

-1.09

 mój

-1.08

 gekenn

-1.08

 habia

-1.08

 moje

-1.08

POSITIVE LOGITS

it

1.55

1.52

()

1.22

Ҭ

1.21

}\

1.20

 заряд

1.16

this

1.13

()}

1.12

),

1.11

to

1.10

Activations Density 0.025%