INDEX

Explanations

the following common nouns

The main thing this neuron does is detect occurrences of the definite article “the,” especially sentence-initial “The.”

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

 alltid

-1.33

GroupLayout

-1.21

éraux

-1.20

 vinyle

-1.16

 епи

-1.16

rierung

-1.10

着他

-1.09

 comté

-1.09

 dîner

-1.09

ジェクト

-1.08

POSITIVE LOGITS

to

1.64

 that

1.27

 choice

1.20

 then

1.18

use

1.17

 Jüng

1.16

 saying

1.14

ないです

1.10

 role

1.10

 examples

1.07

Activations Density 0.065%