INDEX

Explanations

how to instructions

This neuron activates on promotional or sales‐pitch language—especially call-to-action verbs and phrases like “discover,” “learn,” “get” and accompanying punctuation (e.g. colons) used in marketing blurbs.

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

had

-1.03

何か

-1.02

しくお願い

-0.98

 usually

-0.98

比べると

-0.96

upné

-0.96

 estekak

-0.96

🔊

-0.96

lloworld

-0.95

 seemed

-0.94

POSITIVE LOGITS

how

2.59

 your

2.03

you

1.95

why

1.68

 ways

1.52

how

1.29

way

1.23

 genauso

1.23

 strategies

1.23

 jūsų

1.17

Activations Density 0.030%