INDEX

Explanations

numbers twenty to seventy

The neuron flags boilerplate metadata tokens—e.g. version numbers, update dates, license terms, and other header/include keywords—in technical or code‐like text.

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

 FIVE

-0.87

FIVE

-0.82

ால்

-0.72

۲

-0.70

Matin

-0.70

erii

-0.69

THIRD

-0.68

۳

-0.68

 THREE

-0.68

allt

-0.68

POSITIVE LOGITS

 forty

1.16

 fifty

1.12

 sixty

1.09

 Forty

1.00

Thirty

0.98

Forty

0.96

 Thirty

0.96

 twenty

0.96

Fifty

0.95

 seventy

0.93

Activations Density 0.189%