INDEX

Explanations

operating system versions

The neuron strongly activates on tokens made up of multi‐digit numbers (e.g. version numbers, IP segments, status codes).

New Auto-Interp

Configuration

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Negative Logits

 that

-1.71

its

-1.55

of

-1.47

 would

-1.41

 will

-1.39

in

-1.37

 there

-1.37

 could

-1.29

 during

-1.27

if

-1.21

POSITIVE LOGITS

튿

1.31

ｒ

1.27

牙齿

1.16

 bluz

1.14

EARCH

1.13

୬

1.13

1.12

 Cerebral

1.09

 recente

1.09

 terminer

1.09

Activations Density 0.005%