INDEX

Explanations

information about types and links

This neuron activates on words or tokens in foreign languages—i.e. non‐English or non‐basic‐Latin script or diacritic‐marked terms.

New Auto-Interp

Configuration

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Negative Logits

<bos>

-1.30

 콘텐츠

-1.12

<0xA3>

-1.12

 vart

-1.08

捩

-1.05

of

-1.02

かで

-1.02

一时间

-1.01

 düğün

-1.01

 Прези

-1.00

POSITIVE LOGITS

“

1.48

 reikia

1.34

場合は

1.31

you

1.27

：「

1.20

 nėra

1.20

 gjelder

1.16

 tilgjenge

1.16

 naudo

1.14

 zwykle

1.13

Activations Density 0.003%