INDEX

Explanations

author names and dates

The neuron activates on author/creator metadata lines in source‐code comments or docblocks (e.g. “@author X” or “Created by Y”).

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

and

-1.75

:{

-1.27

my

-1.24

 naast

-1.22

meja

-1.18

 avión

-1.18

oaks

-1.18

Ū

-1.18

萬

-1.17

ųjų

-1.16

POSITIVE LOGITS

Porque

1.20

 schuhe

1.20

Елена

1.18

 fasi

1.18

Lieferumfang

1.17

Deine

1.16

 verwendeten

1.16

 despi

1.14

 veik

1.13

 réinitial

1.12

Activations Density 0.013%