INDEX

Explanations

hex codes and sequences

The neuron detects numeric literals—especially floating‐point numbers with decimal points.

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

 использовании

-1.80

𓆉

-1.76

𞤤

-1.75

嫱

-1.72

⚢

-1.67

 asiático

-1.66

擻

-1.65

čiau

-1.64

用到

-1.63

perusahaan

-1.63

POSITIVE LOGITS

at

1.47

how

1.40

its

1.40

蓉

1.39

The

1.34

by

1.33

的血

1.31

 during

1.30

 while

1.30

ваги

1.30

Activations Density 0.003%