INDEX

Explanations

case numbers and dates

The neuron is detecting sequences of digits (often mixed with punctuation) that form identifiers like docket numbers, grant/project codes, or section/subchapter numbers.

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

引用元

-1.53

 schim

-1.46

accor

-1.44

 masina

-1.42

stunning

-1.40

Insee

-1.40

animo

-1.37

 ajudam

-1.33

inverno

-1.33

 essenciais

-1.32

POSITIVE LOGITS

“

1.69

↵

1.52

–

1.51

↵↵

1.39

on

1.38

 more

1.34

du

1.28

1.27

of

1.27

Activations Density 0.273%