INDEX

Explanations

foreign names and places

The neuron fires on capitalized tokens that are parts of names or acronyms—i.e. proper‐noun sequences identifying people, places, organizations, or branded entities.

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

 especially

-1.62

 also

-1.52

 with

-1.52

 actually

-1.51

and

-1.48

 including

-1.45

has

-1.43

 then

-1.37

 this

-1.34

had

-1.34

POSITIVE LOGITS

 ซึ่ง

1.51

 annon

1.44

と呼ばれる

1.41

჻

1.41

 toman

1.33

 což

1.31

뿡

1.30

 pronon

1.30

ཿ

1.30

 fréqu

1.28

Activations Density 0.144%