INDEX

Explanations

Roman numerals followed by periods

This neuron fires strongly on structural or metadata tokens—things like section headings, labels, and other document‐format markers (e.g. “Abstract:”, “Blog”, bullet dots, header numbers, 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

Jäh

-1.67

 někter

-1.62

 niektor

-1.57

Nachteile

-1.54

lorin

-1.52

嵎

-1.52

 kartą

-1.52

érica

-1.51

瞜

-1.48

ικών

-1.47

POSITIVE LOGITS

is

1.87

<em>

1.74

1.59

他是

1.40

 avesse

1.36

<strong>

1.33

↵↵

1.31

 effectuer

1.28

 étages

1.27

if

1.27

Activations Density 0.231%