INDEX

Explanations

questions and specific entities

The neuron fires most strongly on hyperbolic quantifiers and numeric superlatives (e.g. “zillion,” “options,” “every,” “shopper”) typical of promotional, marketing copy.

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

 three

-1.98

two

-1.88

 four

-1.71

 eight

-1.66

 both

-1.66

 five

-1.64

 their

-1.55

 only

-1.52

it

-1.49

my

-1.49

POSITIVE LOGITS

 REALLY

1.34

 našich

1.30

铊

1.29

 delightfully

1.28

après

1.27

 ludicrous

1.27

wiście

1.27

 COULD

1.26

Después

1.25

 quintessential

1.23

Activations Density 0.002%