INDEX

Explanations

characteristic of

The neuron fires on numeric “fraction” tokens that come from PDF‐style page‐count markers (e.g. the “79” in “79 of 144,” “94” in “94 of 144,” etc.).

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

</strong>

-2.05

flä

-1.40

 kön

-1.38

なんですが

-1.30

 sebagainya

-1.29

ᐠ

-1.27

 whose

-1.27

 哪些

-1.25

 whom

-1.24

赓

-1.23

POSITIVE LOGITS

of

2.55

They

1.80

1.58

Your

1.55

д

1.53

 spowod

1.51

1.46

 They

1.45

 them

1.44

ya

1.44

Activations Density 0.087%