INDEX

Explanations

qualities and attributes

This neuron fires on descriptive content words—especially multi-syllabic adjectives and nouns that convey features, qualities, problems, or evaluations.

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

of

-2.28

 from

-2.16

</i>

-1.88

all

-1.86

as

-1.81

for

-1.79

on

-1.78

our

-1.52

not

-1.51

 notable

-1.51

POSITIVE LOGITS

 BEAUTIFUL

1.55

 Fácil

1.45

 simplif

1.42



1.41

 YellowPages

1.40

 manteiga

1.38

 automó

1.36

orters

1.34

 esattamente

1.33

 then

1.30

Activations Density 0.111%