INDEX

Explanations

printing text and errors

The neuron strongly activates on words related to printing and typesetting (e.g., “type,” “typesetting,” “typographical,” “printing,” 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

ご了承

-1.55

(['/

-1.43

//

-1.39

usza

-1.34

This

-1.32

вся

-1.30

vivo

-1.25



-1.25

 wich

-1.24

就没

-1.24

POSITIVE LOGITS

 tých

1.52

 örnek

1.52

 detayl

1.46

 bordados

1.45

 uwag

1.45

 solchen

1.40

 destinada

1.40

Encourage

1.39

 abusos

1.37

榖

1.37

Activations Density 0.054%