INDEX

Explanations

installing, creating, removing

The neuron fires strongly on sentences that issue policy‐style questions or directives (e.g. “What should the government do?” or “Try to find out if…”).

New Auto-Interp

Configuration

Prompts (Dashboard)

392,802 prompts, 256 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Not in Any Lists

Negative Logits

 viciss

0.59

 graced

0.55

 greatly

0.54

 fluctuations

0.54

 possessed

0.54

 adorned

0.54

 olduk

0.53

 contemporaries

0.53

 võivad

0.53

 delighted

0.52

POSITIVE LOGITS

 начать

1.38

 tehdä

1.36

 create

1.35

 remove

1.30

 creare

1.29

 eliminate

1.27

 Remove

1.26

 создать

1.23

 گرفتن

1.23

 Create

1.22

Activations Density 0.556%