INDEX

Explanations

good thing

The neuron fires strongly on words that occur in titles or headings (i.e. section or article headlines).

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

atized

-1.47

 agujero

-1.44

 SPECIFIC

-1.41

ın

-1.39

 watu

-1.38

嵓

-1.38

 insom

-1.34

 colchón

-1.34

 effectuer

-1.32

 différente

-1.31

POSITIVE LOGITS

or

1.64

to

1.62

1.55

 технологі

1.44

at

1.41

 with

1.40

 unruly

1.31

as

1.28

1.27

$\

1.26

Activations Density 0.166%