INDEX

Explanations

The neuron strongly activates on type‐related tokens in code—keywords and identifiers that name types (e.g. “int,” “float,” “void,” single‐letter class names like “B,” “C,” and typedef’d handles).

programming language syntax

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

 energii

-1.59

 biela

-1.53

 Tienen

-1.45

 baskılı

-1.45

 disponibilités

-1.42

 QUÉ

-1.42

cribir

-1.41

＝＝＝＝

-1.40

 esponja

-1.38

moiselle

-1.36

POSITIVE LOGITS

in

1.36

なども

1.27

to

1.21

などは

1.20

ora

1.19

が

1.16

でも

1.14

 even

1.13

 incluso

1.12

 heridos

1.11

Activations Density 0.002%