INDEX

Explanations

code and structured data

The neuron strongly activates on isolated numeric tokens (e.g. version numbers, figure or equation labels) rather than ordinary 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

«

-0.97

tâ

-0.95

 désigner

-0.91

 représenter

-0.89

saison

-0.88

 tør

-0.88

 révélé

-0.88

妡

-0.87

 thom

-0.87

riculares

-0.86

POSITIVE LOGITS

ﾐ

1.03

ヒント

0.88

"",

0.88

Matéri

0.88

ギフト

0.87

 cursing

0.87

ResponseType

0.86

 Redacción

0.85

 fattori

0.85

λεκ

0.84

Activations Density 0.047%