INDEX

Explanations

archive/pre-print identifiers, place names, stock tickers

The neuron fires on technical identifier tokens—things like arXiv categories (e.g. “hep-ph”), document‐lab codes or report numbers (e.g. “SU-F-BRCD-04”), project or stock tickers (e.g. “PEUP.PA”), and similar alphanumeric labels.

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

it

-1.59

 used

-1.41

 during

-1.34

 while

-1.34

the

-1.27

 which

-1.25

 throughout

-1.21

 piratas

-1.20

 before

-1.20

 suro

-1.20

POSITIVE LOGITS

这种

1.59

тері

1.52

ана

1.30

写真

1.27

 лишь

1.27

abonnement

1.25

 pertes

1.25

ómico

1.24

ONLY

1.23

Этот

1.23

Activations Density 0.008%