INDEX

Explanations

introduces a new characteristic

The neuron fires on evaluative degree words—especially adjectives and adverbs that compare or emphasize (e.g. “more,” “better,” “faster,” “essential,” “simplified”).

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

to

-1.42

ni

-1.41

</strong>

-1.31

-1.30

at

-1.27

-1.25

-1.20

ong

-1.20

te

-1.18

ut

-1.18

POSITIVE LOGITS

ühnen

1.33

 tracción

1.30

ższych

1.28

vosť

1.27

 daje

1.27

 would

1.25

 nastro

1.22

 所以

1.21

umo

1.20

 proveniente

1.20

Activations Density 0.067%