INDEX

Explanations

dog tags, ITS, shrouds, British, MS surfaces

The neuron fires on content-heavy nouns or noun compounds—especially technical or dramatic descriptor terms—highlighting key topic words.

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

 letto

-1.52

if

-1.51

 minore

-1.47

 when

-1.47

 setengah

-1.34

 fondu

-1.34

 abitanti

-1.32

𝘥

-1.31

 crede

-1.29

鈦

-1.29

POSITIVE LOGITS

return

1.59

</em>

1.43

ziest

1.40

 počíta

1.37

 nabíd

1.36

消失了

1.36

leggen

1.30

heizung

1.26

 }}">

1.24

adaptiveStyles

1.24

Activations Density 0.085%