INDEX

Explanations

verbs of speech and thought

The main thing this neuron does is detect speaker-attribution phrases (e.g. “he said,” “they say”).

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

at

-2.02

in

-1.84

 because

-1.80

 their

-1.59

 Both

-1.54

 just

-1.53

 Their

-1.50

 even

-1.45

on

-1.44

Siapa

-1.44

POSITIVE LOGITS

还不错

1.66

۲

1.64

 perioden

1.57

 که

1.55

۳

1.53

ʻ

1.52

可以

1.52

рации

1.50

所有

1.48

 provenant

1.48

Activations Density 0.017%