INDEX

Explanations

numbers followed by characters

This neuron fires on numeric tokens—especially the digit “3”—when they appear as part of version numbers, measurements, or other numeric data.

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

 після

-1.39

 after

-1.36

에서

-1.35

 before

-1.32

理由は

-1.31

 While

-1.30

するために

-1.26

陕

-1.23

 With

-1.23

).

-1.23

POSITIVE LOGITS

 aprob

1.45

蛲

1.45

simply

1.34

 geheim

1.33

 probablement

1.30

 konkret

1.28

 этим

1.27

alarme

1.27

 malen

1.25

 redus

1.25

Activations Density 0.075%