INDEX

Explanations

author, abstract, findings, redistribute

This neuron detects the decorative structure and metadata tags (e.g. the repeated “*” lines and “@file,” “@author,” “@version,” etc.) of C-style header comment blocks.

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

 dvě

-1.43

 mismas

-1.24

cách

-1.23

 everything

-1.22

 joyas

-1.20

很快

-1.20

 deux

-1.19

 criaturas

-1.19

 Bagaimana

-1.19

 levanta

-1.18

POSITIVE LOGITS

靭

1.34

かは

1.32

1.30

 ribu

1.25

 marts

1.21

 will

1.16

of

1.13

靱

1.12

 intero

1.10

看去

1.10

Activations Density 0.091%