INDEX

Explanations

need or want assistance

The neuron fires on words expressing user desires or prompts—especially “want” and “need” (and related phrasing asking if you “don’t want” or “not sure”) in calls to action or questions.

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

黉

-1.55

 milj

-1.52

 даже

-1.45

是可以

-1.38

joueurs

-1.37

 hemd

-1.37

亹

-1.35

ленно

-1.34

 banyo

-1.33

遢

-1.31

POSITIVE LOGITS

or

1.97

in

1.90

 wants

1.59

 makes

1.56

put

1.47

 when

1.46

 gets

1.45

you

1.45

 like

1.44

 find

1.42

Activations Density 0.054%