INDEX

Explanations

links to webpages

The neuron strongly activates on URL fragments (e.g. domain names and slashes), i.e. it detects parts of web links.

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

着一个

-1.59

냑

-1.45

to

-1.39

 WHEN

-1.32

摸了

-1.30

 idyllic

-1.29

۴

-1.27

 wist

-1.27

‌

-1.27

一座

-1.27

POSITIVE LOGITS

our

1.47

 their

1.41

 other

1.41

！！”

1.41

 multitude

1.40

 from

1.36

 dejar

1.36

 possibly

1.34

یشن

1.34

breasted

1.30

Activations Density 0.022%