INDEX

Explanations

lists of rules

The neuron fires on content‐heavy, information‐rich tokens—like key nouns, names, and technical terms—that carry the core facts of a news or report-style sentence.

New Auto-Interp

Configuration

Prompts (Dashboard)

392,802 prompts, 256 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Not in Any Lists

Negative Logits

 এরূপ

1.10

このような

1.05

 нашего

1.02

 हूँ

1.01

 এইরূপ

0.98

த்தினை

0.97

cca

0.96

 моего

0.92

においても

0.92

 данного

0.89

POSITIVE LOGITS

 staffers

1.27

 lawmakers

1.16

 officials

1.13

 backers

1.13

 dozens

1.09

 apiece

1.04

 amid

1.02

 prosecutors

1.02

—

1.01

 overseen

1.00

Activations Density 1.369%