INDEX

Explanations

The neuron is acting as a “progress” detector—its activation grows the farther into the document a token appears.

Furtherance or addition

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

 Terkait

-1.24

 хорошие

-1.23

 моей

-1.23

 monté

-1.20

often

-1.19

 mojej

-1.17

artiges

-1.17

bedingt

-1.16

 cewek

-1.15

 bluz

-1.15

POSITIVE LOGITS

 your

1.98

or

1.86

you

1.85

さらに

1.82

get

1.63

更に

1.45

 also

1.37

 truly

1.27

 Then

1.24

Then

1.24

Activations Density 0.033%