INDEX

Explanations

descriptive qualities and states

The neuron fires on descriptive, evaluative language—adjectives and the key nouns they modify when characterizing people or items (e.g. outgoing, shy, film, property, truck) in reviews and personal descriptions.

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

or

-1.66

 when

-1.53

on

-1.48

in

-1.44

 this

-1.43

 during

-1.37

 before

-1.36

 through

-1.32

 because

-1.30

 instead

-1.26

POSITIVE LOGITS

 Incluso

1.28

 with

1.25

 المثال

1.13

anhas

1.13

 linken

1.10

 โดย

1.09

 Tienen

1.09

椋

1.09

粕

1.07

俵

1.05

Activations Density 0.040%