INDEX

Explanations

getting started

The neuron fires on the phrase “get started,” detecting instructions or prompts that guide the reader on how to begin doing something.

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

还真

-1.26

 grij

-1.25

uye

-1.23

 opdracht

-1.15

 情報

-1.13

Meaning

-1.13

ってない

-1.13

黉

-1.09

 dochter

-1.09

厲害

-1.08

POSITIVE LOGITS

 start

1.89

 with

1.80

start

1.51

 begin

1.48

بدأ

1.39

rupal

1.31

任何

1.28

ांकि

1.26

 откуда

1.26

 begins

1.24

Activations Density 0.022%