INDEX

Explanations

first person statements

The neuron fires on first‐person self-references—tokens like “I,” “my,” “have,” “been,” etc., in personal statements.

New Auto-Interp

Configuration

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Negative Logits

歺

-1.69

 yetu

-1.54

 aquilo

-1.54

 papiers

-1.52

){$

-1.50

 blusa

-1.48

 comigo

-1.48

 sowieso

-1.48

脔

-1.46

 този

-1.45

POSITIVE LOGITS

to

2.78

at

1.98

if

1.81

as

1.63

now

1.55

 work

1.52

 able

1.52

in

1.52

of

1.49

 provide

1.48

Activations Density 0.037%