INDEX

Explanations

proposing actions or changes

This neuron fires on action-oriented words that introduce the aims or proposals of a research plan (e.g. verbs like “organize,” “create,” “propose,” “use,” “apply”).

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

 этих

-1.66

 these

-1.51

CONCLUSIONES

-1.39

 diesen

-1.30

 rassemb

-1.27

 dessas

-1.25

 destes

-1.23

these

-1.23

//!

-1.20

 comparação

-1.18

POSITIVE LOGITS

if

1.90

but

1.83

by

1.72

1.70

 without

1.68

or

1.58

 should

1.58

as

1.50

 using

1.42

 options

1.42

Activations Density 0.119%