INDEX

Explanations

items coupled by 'and'

The main thing this neuron does is detect numeric tokens (e.g. years, version numbers, counts, and other digits).

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

}$.

-1.63

。

-1.52

،

-1.48

etc

-1.41

などなど

-1.40

 poderão

-1.39

 поэтому

-1.38

など

-1.37

 even

-1.37

-1.36

POSITIVE LOGITS

and

3.73

 וגם

1.87

AND

1.86

 maupun

1.58

性和

1.57

 lẫn

1.38

ecuted

1.38

panjang

1.32

 сегодняш

1.32

casse

1.31

Activations Density 0.252%