INDEX

Explanations

code structure and foreign words

The neuron fires on numeric literals in code (integer constants).

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

for

-1.68

 just

-1.55

 hunde

-1.54

 could

-1.46

"),

-1.45

ৄ

-1.45

if

-1.44

 only

-1.38

häuse

-1.37

 向け

-1.37

POSITIVE LOGITS

 astounding

1.41

 insanely

1.41

and

1.39

 unbelievably

1.34

 exceedingly

1.34

 terrific

1.34

 diversos

1.34

 appunto

1.31

 sobretudo

1.30

 delightfully

1.29

Activations Density 0.019%