INDEX

Explanations

boolean attributes

The neuron fires primarily on affirmative boolean attribute values—especially the token “YES” (and similar “true”/“yes”) in XML or config files.

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

of

-1.73

 from

-1.69

 their

-1.63

 that

-1.63

 either

-1.59

 entweder

-1.57

 know

-1.42

 musíte

-1.34

 understand

-1.34

 considered

-1.34

POSITIVE LOGITS

 vides

1.43

 conten

1.36

 moderne

1.34

 comte

1.32

viv

1.32

 capa

1.32

 thèse

1.31

uteurs

1.30

 langue

1.28

 confor

1.28

Activations Density 0.047%