INDEX

Explanations

device which, attr LIKE

The neuron fires on tokens that look like code or schema identifiers (e.g. variable/field names such as “columns,” “device,” “attr,” or other named symbols).

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

 eliminates

-1.47

it

-1.47

as

-1.45

 that

-1.43

 with

-1.38

two

-1.34

 when

-1.34

 provides

-1.32

 using

-1.32

or

-1.32

POSITIVE LOGITS

 whimsical

1.31

 eene

1.30

 monstrous

1.29

 ſen

1.28

 gratuite

1.25

 prenn

1.21

 grotte

1.19

 perfecte

1.19

 goud

1.18

抠

1.17

Activations Density 0.020%