INDEX

Explanations

numbers and citations in brackets

The neuron fires on numeric tokens—especially reference/index numbers, years, page or figure numbers, and other numeric metadata in scholarly or structured 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

is

-1.43

-1.16

-1.10

-1.05

has

-1.02

летов

-0.99

还得

-0.98

ang

-0.97

ag

-0.96

POSITIVE LOGITS

 biela

1.18

 primeiras

1.14

tamment

1.12

Seni

1.09

daj

1.06

 école

1.05

broj

1.03

Penjelasan

1.03

 ciclismo

1.02

Tú

1.02

Activations Density 0.003%