INDEX

Explanations

class, file, identifier, system, reflection

The neuron fires on short acronym‐style tokens (one to three letters long, often all‐caps or mixed‐case) such as “TJ,” “Tb,” “Os,” “Sc,” “So,” “Kt,” or “Co.”

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

 they

-1.68

蜮

-1.52

 něj

-1.48

 reunió

-1.45

惬

-1.45

контро

-1.42

 コーデ

-1.42

 with

-1.37

ваш

-1.35

ovka

-1.34

POSITIVE LOGITS

什么样的

1.56

で見る

1.48

Actually

1.45

getF

1.45

However

1.37

any

1.34

綃

1.33

ഗ

1.32

since

1.30

Activations Density 0.088%