INDEX

Explanations

research topics and outputs

The neuron fires on formal research-context terminology—like document headings and administrative terms (e.g. “Research,” “Facilities,” “proposals,” “review,” “Center,” “domain projects,” 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

我会

-1.67

はコチラ

-1.64

'>

-1.61

',

-1.61

 semanas

-1.59

 เป็น

-1.55

idemia

-1.52

 roast

-1.50

ztes

-1.48



-1.48

POSITIVE LOGITS

it

1.54

it

1.52

/'.$

1.45

⸰

1.44

 that

1.44

 migliori

1.43

knię

1.40

 soste

1.37

 singola

1.36

and

1.35

Activations Density 0.076%