INDEX

Explanations

programming and system paths

The neuron fires on distinctive multi-character identifier tokens—things like module or library names, function names, variable names, or domain names in code rather than everyday words.

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

 other

-1.52

 even

-1.40

 without

-1.34

any

-1.27

 before

-1.25

ॄ

-1.24

all

-1.22

 wakil

-1.20

 always

-1.19

 deviceId

-1.16

POSITIVE LOGITS

 vaše

1.43

 naše

1.33

ブラシ

1.32

ल्ड

1.30

 жк

1.28

hörer

1.27

まれた

1.27

mmation

1.27

PushMatrix

1.26

युक्त

1.22

Activations Density 0.001%