INDEX

Explanations

invention relates to

The neuron fires on words and short phrases that signal emphasis or importance (e.g. “excellent,” “especially,” “Most importantly,” “needs to,” “I am not sure”).

New Auto-Interp

Configuration

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Negative Logits

 eventuell

-1.63

 maybe

-1.52

 also

-1.48

 البته

-1.46

but

-1.46

自体は

-1.42

感じが

-1.35

のかもしれません

-1.35

 talvez

-1.34

-1.33

POSITIVE LOGITS

 these

1.51

 those

1.45

 three

1.37

是

1.34

»?

1.31

 will

1.26

 sayesinde

1.24

»:

1.23

 because

1.23

 подчер

1.22

Activations Density 0.125%