INDEX

Explanations

code followed by identifier

The neuron flags numeric tokens—especially numbers (including decimals and negatives) used in product codes, quantities, and specifications.

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

-1.23

⌬

-1.09

象徴

-1.09

 problemas

-1.05

你可以

-1.05

構造

-1.04

 politische

-1.04

 Probleme

-1.04

fetched

-1.03

 montagna

-1.03

POSITIVE LOGITS

пку

1.28

to

1.23

of

1.21

 Пусть

1.17

médiaire

1.17

code

1.16

 verändert

1.16

𝚃

1.16

msch

1.16

 już

1.15

Activations Density 0.026%