INDEX

Explanations

conventional rectilinear

The neuron fires on multi‐syllabic, content‐bearing words (i.e. non–function words—technical, topic‐specific or otherwise “meaningful” terms) rather than on common grammatical tokens.

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

 will

-1.22

 because

-1.18

when

-1.17

 create

-1.15

 epitom

-1.14

On

-1.13

 with

-1.13

 There

-1.12

As

-1.12

can

-1.12

POSITIVE LOGITS

mainly

1.45

luy

1.45

basically

1.45

 gewi

1.43

 verhind

1.41

蚡

1.35

 bezw

1.35

 spektak

1.33

 bụ

1.31

 SAYS

1.30

Activations Density 0.086%