INDEX

Explanations

company names and lists

The neuron strongly activates on subword tokens that mark the beginnings of multi‐character words (i.e. the initial BPE pieces of longer words).

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

in

-1.50

 dernière

-1.27

 Codable

-1.26

格好

-1.25

we

-1.24

家は

-1.22

 adecu

-1.20

 buque

-1.20

 inici

-1.19

取り組み

-1.19

POSITIVE LOGITS

""

1.34

▪

1.28

➢

1.23

''

1.22

晶

1.21

 alcune

1.20

</em>

1.20

and

1.19

?,

1.16

 conglomer

1.15

Activations Density 0.006%