INDEX

Explanations

numbers, letters, and codes

The neuron fires on tokens that contain digits (e.g. standalone numbers, numbers with punctuation, or alphanumeric labels with 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

as

-1.24

 also

-1.17

 when

-1.17

 When

-1.13

of

-1.11

not

-1.09

 will

-1.05

 after

-1.05

 instead

-1.04

for

-1.02

POSITIVE LOGITS

 många

1.05

 rami

1.04

︀

1.03

Produk

1.02

andoah

1.01

 épar

1.00

 Samoa

0.99

 ועוד

0.98

ómicos

0.97

тельством

0.96

Activations Density 0.019%