INDEX

Explanations

actions and their outcomes

This neuron fires on words marking reported events or actions (especially past‐tense verbs and their nominalizations, e.g. “raised,” “issued,” “granted,” “unveiled,” “announcement”).

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

 eksemp

-1.58

 chacune

-1.53

 nécessairement

-1.45

STERY

-1.41

 conséquent

-1.41

欧

-1.40

 caufe

-1.37

相比

-1.32

ズニー

-1.32

reservations

-1.31

POSITIVE LOGITS

at

1.52

[];

1.47

);

1.38

you

1.38

common

1.37

 your

1.32

1.31

to

1.30

are

1.27

iy

1.27

Activations Density 0.042%