INDEX

Explanations

web links

The neuron fires on structural or formatting tokens that mark text boundaries or omissions—e.g. bracketed ellipses “[…]”, quotation-block “>” markers, inline metadata like “wet-web” headings, 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

、【

-1.61

-1.59

デメリット

-1.46

してみると

-1.45

 terbesar

-1.45

 chère

-1.43

 klachten

-1.43

 gracioso

-1.41

Ⱥ

-1.41

 réus

-1.40

POSITIVE LOGITS

</i>

1.75

这个

1.56

1.53

gucig

1.52

這

1.52

既然

1.51

</h2>

1.50

还是

1.49

}$.

1.48

Activations Density 0.002%