INDEX

Explanations

name and place continuations, code prefixes

This neuron detects headings or titles, such as table header tags or dateline/section labels often shown in all-caps or title case.

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

 мы

-2.02

-1.95

</em>

-1.95

 venant

-1.83

stalt

-1.83

-1.78

<em>

-1.77

 такой

-1.75

 Zustimmung

-1.73

 تصریح

-1.73

POSITIVE LOGITS

2.44

 sabemos

1.92

</h1>

1.86

 they

1.81

 kteří

1.68

ющие

1.66

 jargon

1.65

 involucra

1.65

porque

1.64

</th>

1.63

Activations Density 0.000%