INDEX

Explanations

news headlines, legal filings

The neuron activates on high‐contrast document header and metadata tokens—things like all-caps labels, volume numbers, dates, and other structured heading elements.

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

 your

-2.06

 شما

-1.63

our

-1.40

 youre

-1.39

now

-1.38

 você

-1.34

 вам

-1.29

你也

-1.27

CÓ

-1.27

귀

-1.27

POSITIVE LOGITS

啻

1.44

 ILLINOIS

1.41

癡

1.39

 them

1.37

 portatil

1.34

 δύο

1.33

 tdci

1.30

REUTERS

1.27

🤣🤣

1.26

始まる

1.23

Activations Density 0.175%