INDEX

Explanations

according to a source

The neuron fires on source‐attribution phrases—especially “According to …” constructions and the names or organizations that follow.

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

in

-2.39

for

-2.19

and

-1.77

on

-1.70

 from

-1.55

 into

-1.49

 without

-1.47

or

-1.41

at

-1.41

 Each

-1.38

POSITIVE LOGITS

鹟

1.63

 percor

1.55

 テープ

1.55

 pérd

1.54



1.51

☆

1.49

stalgia

1.48

秾

1.48

 interrom

1.46

蒌

1.46

Activations Density 0.026%