INDEX

Explanations

conjunctions introducing contrast or addition

The neuron strongly fires on common discourse/connective words (e.g. “but,” “also,” “before,” “those,” “who,” “to,” etc.), i.e. transition or conjunction tokens.

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

ссия

-1.09

を持っている

-1.08

 rất

-1.00

 keten

-0.99

мата

-0.98

ében

-0.98

 voler

-0.98

めんなさい

-0.97

 there

-0.94

 nostre

-0.94

POSITIVE LOGITS

if

1.83

 your

1.52

不要

1.40

 если

1.36

 jeśli

1.29

 should

1.26

you

1.21

 yourself

1.18

be

1.17

 keep

1.11

Activations Density 0.029%