INDEX

Explanations

linking aadhar

This neuron fires strongly on tokens that include the unusual “ ” (unicode U+2004) padding (i.e. non-ASCII whitespace) rather than ordinary text.

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

 великолеп

-1.51

brille

-1.29

معرفی

-1.23

schuhe

-1.22

zg

-1.20

Manfaat

-1.18

langsung

-1.17

ثيق

-1.17

 naves

-1.17

wunsch

-1.16

POSITIVE LOGITS

1.30

—

1.27

才可以

1.22

 updatedAt

1.20

preneurship

1.17

就行

1.15

Comté

1.13

Activations Density 0.038%