INDEX

Explanations

leading words or symbols

This neuron is triggered by stray or non-standard punctuation and symbol fragments (the little bits of “gibberish” like isolated tildes, odd letters, ligature fragments, and PDF-artifact symbols).

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

and

-1.78

 noteworthy

-1.56

 ingrained

-1.54

now

-1.40

 financial

-1.39

 pertaining

-1.39

Lepid

-1.38

 staunch

-1.37

 multifaceted

-1.35

 good

-1.34

POSITIVE LOGITS

!!!!!!!

1.34



1.34

entire

1.30

cai

1.28

 !!!!

1.25

 câbles

1.24

commonly

1.23

both

1.23

 приготовления

1.23

exactly

1.22

Activations Density 0.006%