INDEX

Explanations

The neuron fires on tokens that occur in music‐release metadata—dates, numbers, and key credit words (e.g. “recorded,” “by,” “drop,” artist/album names)—i.e. information about when and by whom a song or album was released.

song title or artist

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

at

-1.46

 several

-1.44

 three

-1.43

 znaleźć

-1.37

 when

-1.36

 even

-1.34

 five

-1.27

kinh

-1.23

 some

-1.23

two

-1.23

POSITIVE LOGITS

 vider

1.44

 herbes

1.41

 nive

1.34

 tene

1.30

 intrig

1.30

 pengar

1.29

 gesta

1.29

oolean

1.27

 vitamine

1.26

菒

1.26

Activations Density 0.003%