INDEX

Explanations

numerical citations

This neuron selectively activates on the numeric and punctuation tokens used in academic reference metadata (volume, issue, page numbers, DOIs, 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

in

-2.34

and

-2.20

by

-2.08

on

-2.03

or

-1.82

 here

-1.57

 like

-1.55

 bahkan

-1.52

})(\

-1.51

).

-1.50

POSITIVE LOGITS

 certaines

1.84

︐

1.83

набор

1.78

ཥ

1.70

 vengono

1.65

⅏

1.63

 suffisamment

1.62

Ꮠ

1.60

 vieilles

1.58

bró

1.56

Activations Density 0.026%