INDEX

Explanations

news publications

The neuron fires on the names of news outlets or media publications.

New Auto-Interp

Configuration

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Negative Logits

to

-1.67

for

-1.58

 without

-1.28

 that

-1.23

as

-1.22

 when

-1.21

at

-1.20

 from

-1.20

 because

-1.18

-1.16

POSITIVE LOGITS

 parut

1.35

 geene

1.31

Ꮴ

1.30

manufactured

1.27

 jaket

1.25

 mérit

1.25

 cliquer

1.23

嘁

1.23

 discuter

1.22

 longues

1.21

Activations Density 0.041%