INDEX

Explanations

words following commas

The neuron is sensitive to capitalized technical terms and acronyms (domain-specific buzzwords).

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

for

-1.44

=[];

-1.30

 when

-1.29

 papiers

-1.24

);*/

-1.23

した後

-1.17

 groupName

-1.16

llte

-1.16

ness

-1.16

➗

-1.14

POSITIVE LOGITS

 проблему

1.38

indro

1.23

agic

1.23

 preva

1.23

zku

1.21

我没有

1.20

ICIENCY

1.19

burton

1.17

 baud

1.16

1.15

Activations Density 0.181%