INDEX

Explanations

years and roman numerals

The neuron strongly activates on numeric tokens representing years (four-digit date 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

夲

-1.73

 मिन

-1.58

Whoever

-1.58

 هنوز

-1.55

pios

-1.54

Quieres

-1.54

🩱

-1.52

 favoritas

-1.52

ᐅ

-1.51

 dîner

-1.50

POSITIVE LOGITS

the

1.83

</h3>

1.60

and

1.48

to

1.45

1.42

 three

1.41

as

1.41

an

1.41

1.40

Activations Density 0.035%