INDEX

Explanations

[@ citations

The neuron specifically picks out numeric tokens, in particular the four-digit year (or other numbers) appearing in citations and reference lists.

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

if

-1.14

 bude

-1.13

 veliko

-1.05

$.}

-1.05

 That

-1.05

of

-1.00

 ако

-0.99

 каждо

-0.98

有事

-0.98

 всеки

-0.96

POSITIVE LOGITS

oczes

1.00

prende

0.95

Pcs

0.94

',"

0.94

きの

0.94

 thumbnails

0.92

 contextLoads

0.91

体は

0.91

 ..............

0.91

 ścian

0.91

Activations Density 0.002%