INDEX

Explanations

explanation and concepts

The neuron fires most strongly on words that label or introduce structured list entries or section headings (e.g. “Explanation,” “Questions,” “Review,” “Materials,” “Files,” etc.).

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

 this

-1.27

えっ

-1.14

 sinyal

-1.08

 می‌گو

-1.07

Otra

-1.05

を持っている

-1.05

おまけ

-1.05

ffiche

-1.05

ابه

-1.04

commod

-1.03

POSITIVE LOGITS

of

1.59

and

1.52

 including

1.36

 bezüglich

1.18

 hinsichtlich

1.16

۔

1.16

nguyen

1.16

กับ

1.05

,'"

1.04

 calciatore

1.04

Activations Density 0.017%