INDEX

Explanations

words followed by punctuation or certain words

The neuron fires on distinctive content words—especially proper names, product or company names, and technical terms.

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

are

-4.19

is

-3.94

 will

-3.36

 would

-3.16

 should

-2.88

 have

-2.83

was

-2.69

 could

-2.66

may

-2.53

 must

-2.48

POSITIVE LOGITS

llten

1.91

 mengucap

1.88

vergleich

1.86

 dámské

1.80

thanx

1.80

蛱

1.78



1.78

鮡

1.76

HAHA

1.75

Activations Density 1.329%