INDEX

Explanations

Item numbers and names

The neuron flags tokens that are either numeric strings or capitalized identifiers (e.g. digits like “302”, “400”, “0” and proper‐noun/class‐name tokens like “Item” or “John”).

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

and

-1.73

you

-1.66

满脸

-1.60

神色

-1.58

geschäft

-1.48

⽼

-1.45

 нашим

-1.41

 ogran

-1.41

一脸

-1.39

 they

-1.36

POSITIVE LOGITS

虽说

1.73

1.69

 serdecznie

1.59

م

1.50

玉ねぎ

1.48

萏

1.48

Læs

1.48

ories

1.45

 camada

1.44

Activations Density 0.001%