INDEX

Explanations

list of items

The neuron is highly sensitive to very common function words (e.g. “for,” “of,” “is,” “a”) in technical or documentation‐style text.

New Auto-Interp

Configuration

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Negative Logits

 after

-1.55

 before

-1.45

 without

-1.41

or

-1.36

on

-1.25

 because

-1.25

had

-1.23

 during

-1.23

 there

-1.17

 gave

-1.13

POSITIVE LOGITS

 både

1.51

Including

1.40

だけでなく

1.36

 rokok

1.34

だけではなく

1.34

 Inſ

1.29

 geene

1.26

both

1.21

various

1.19

 eksklu

1.18

Activations Density 0.149%