INDEX

Explanations

The neuron is triggered by phrases that point to where a resource can be accessed (e.g. “available at,” “retrieved from”).

at http/https or the

New Auto-Interp

Configuration

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Negative Logits

 arşivlendi

-1.32

で見る

-1.20

They

-1.15

ワ

-1.10

ごちそう

-1.09

Because

-1.09

伀

-1.09

-1.05

PORARY

-1.05

Instead

-1.04

POSITIVE LOGITS

 after

1.70

or

1.43

1.40

 during

1.27

erobic

1.18

at

1.15

 škoda

1.13

 with

1.12

in

1.12

for

1.10

Activations Density 0.004%