INDEX

Explanations

code formatting and completion

The neuron fires on occurrences of the boolean literals “True” and “False” in code.

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

 this

-1.66

of

-1.57

 their

-1.45

ambut

-1.41

at

-1.33

 which

-1.29

 only

-1.27

 would

-1.21

or

-1.20

his

-1.20

POSITIVE LOGITS

 ARISING

1.52

anium

1.48

神

1.38

alers

1.25

offers

1.24

becomes

1.23

ureusement

1.20

enhancing

1.19

continues

1.19

tells

1.19

Activations Density 0.055%