INDEX

Explanations

phrases, questions, and titles

The neuron fires almost exclusively on the very first content word of a quoted UI/localization string (e.g. the first token after the opening quote in translation files).

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

our

-1.68

 this

-1.52

 these

-1.30

 that

-1.26

 Données

-1.24

 Conseils

-1.24

 olvido

-1.22

鷽

-1.21

 lyder

-1.21

 those

-1.16

POSITIVE LOGITS

to

1.56

 ähnlichen

1.09

 يت

1.06

 fremden

1.06

TRIBUN

1.04

was

1.03

 kleineren

1.02

୪

0.99

걍

0.98

 replete

0.98

Activations Density 0.001%