INDEX

Explanations

dateline separators

The neuron fires on wire-service bylines or source‐credit markers (e.g. “PRNewswire/,” “ANI,” “GLOBE NEWS WIRE”) in news‐style text.

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

ྂ

-1.72

ittää

-1.61

ruckt

-1.58

}>{

-1.55

 "]";

-1.55

*{

-1.53

ulfon

-1.53

ib

-1.53

डे

-1.52

irical

-1.51

POSITIVE LOGITS

 with

1.83

Those

1.78

Suerte

1.75

Prueba

1.67

1.66

Knowing

1.65

1.63

we

1.62

1.60

Activations Density 0.074%