INDEX

Explanations

The neuron fires on numeric tokens and sequences (e.g. version numbers, IDs, measurements) in structured or code‐like data.

optical, physics, materials

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

Jeder

-1.29

ּוֹ

-1.27

영어

-1.23

ādā

-1.15

ṓ

-1.14

OGRAPH

-1.12

IActionResult

-1.10

 regeringen

-1.09

magini

-1.09

}>{

-1.09

POSITIVE LOGITS

and

1.46

but

1.41

at

1.27

 левый

1.04

0.99

場面

0.96

ve

0.96

랜드

0.93

Bur

0.92

0.91

Activations Density 0.003%