INDEX

Explanations

United States District Court

This neuron detects formal court-jurisdiction headings in legal documents, especially tokens naming the “United States” courts (e.g. “United States,” “U.S.,” “Court,” “Appeals,” “Circuit”).

New Auto-Interp

Configuration

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Negative Logits

on

-1.95

 with

-1.94

in

-1.55

or

-1.45

 that

-1.36

 metod

-1.30

all

-1.28

and

-1.28

 from

-1.22

 passé

-1.21

POSITIVE LOGITS

 offensichtlich

1.55

 weitgehend

1.47

1.38

 nabíz

1.37

 meistens

1.34

 传统

1.31

 festgelegt

1.28

 begrenzt

1.27

 thirteenth

1.27

 verwal

1.26

Activations Density 0.107%