INDEX

Explanations

how to do

The neuron activates on non-Latin (CJK) character tokens, i.e. words or morphemes in Japanese or Chinese script.

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

ï

-1.59

 ไม่

-1.30

 ��

-1.24

 که

-1.23

 для

-1.23

¦

-1.22

zież

-1.21

 simplemente

-1.20

ж

-1.18

 один

-1.17

POSITIVE LOGITS

or

1.39

Vidéo

1.35

越大

1.35

 nachts

1.34

 négo

1.32

鏵

1.32

 several

1.30

 this

1.28

 Autó

1.28

 morado

1.27

Activations Density 0.011%