INDEX

Explanations

identifying "Background" sections

This neuron identifies section header labels (often in all caps) such as “BACKGROUND:”, “OBJECTIVE:”, “PURPOSE:”, and similar document or code block headers.

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

 disparition

-1.51

摸了

-1.41

心头

-1.37

ミル

-1.34

你们的

-1.30

-1.29

レザー

-1.27

ほか

-1.27

 Konkurrenz

-1.26

 motorbike

-1.23

POSITIVE LOGITS

 bambina

1.55

 用品

1.40

 These

1.38

ഔ

1.33

 նաև

1.31

蹉

1.30

 generazione

1.29

 informed

1.27

inverno

1.26

 hubo

1.22

Activations Density 0.011%